34aa855df732c4ffa35c174ea1ba2a8dc81bbfd9
1/*
2 * Copyright (C) 2005 Junio C Hamano
3 */
4#include "cache.h"
5#include "tempfile.h"
6#include "quote.h"
7#include "diff.h"
8#include "diffcore.h"
9#include "delta.h"
10#include "xdiff-interface.h"
11#include "color.h"
12#include "attr.h"
13#include "run-command.h"
14#include "utf8.h"
15#include "userdiff.h"
16#include "submodule-config.h"
17#include "submodule.h"
18#include "ll-merge.h"
19#include "string-list.h"
20#include "argv-array.h"
21#ifdef NO_FAST_WORKING_DIRECTORY
23#define FAST_WORKING_DIRECTORY 0
24#else
25#define FAST_WORKING_DIRECTORY 1
26#endif
27static int diff_detect_rename_default;
29static int diff_compaction_heuristic; /* experimental */
30static int diff_rename_limit_default = 400;
31static int diff_suppress_blank_empty;
32static int diff_use_color_default = -1;
33static int diff_context_default = 3;
34static const char *diff_word_regex_cfg;
35static const char *external_diff_cmd_cfg;
36static const char *diff_order_file_cfg;
37int diff_auto_refresh_index = 1;
38static int diff_mnemonic_prefix;
39static int diff_no_prefix;
40static int diff_stat_graph_width;
41static int diff_dirstat_permille_default = 30;
42static struct diff_options default_diff_options;
43static long diff_algorithm;
44static char diff_colors[][COLOR_MAXLEN] = {
46 GIT_COLOR_RESET,
47 GIT_COLOR_NORMAL, /* CONTEXT */
48 GIT_COLOR_BOLD, /* METAINFO */
49 GIT_COLOR_CYAN, /* FRAGINFO */
50 GIT_COLOR_RED, /* OLD */
51 GIT_COLOR_GREEN, /* NEW */
52 GIT_COLOR_YELLOW, /* COMMIT */
53 GIT_COLOR_BG_RED, /* WHITESPACE */
54 GIT_COLOR_NORMAL, /* FUNCINFO */
55};
56static int parse_diff_color_slot(const char *var)
58{
59 if (!strcasecmp(var, "context") || !strcasecmp(var, "plain"))
60 return DIFF_CONTEXT;
61 if (!strcasecmp(var, "meta"))
62 return DIFF_METAINFO;
63 if (!strcasecmp(var, "frag"))
64 return DIFF_FRAGINFO;
65 if (!strcasecmp(var, "old"))
66 return DIFF_FILE_OLD;
67 if (!strcasecmp(var, "new"))
68 return DIFF_FILE_NEW;
69 if (!strcasecmp(var, "commit"))
70 return DIFF_COMMIT;
71 if (!strcasecmp(var, "whitespace"))
72 return DIFF_WHITESPACE;
73 if (!strcasecmp(var, "func"))
74 return DIFF_FUNCINFO;
75 return -1;
76}
77static int parse_dirstat_params(struct diff_options *options, const char *params_string,
79 struct strbuf *errmsg)
80{
81 char *params_copy = xstrdup(params_string);
82 struct string_list params = STRING_LIST_INIT_NODUP;
83 int ret = 0;
84 int i;
85 if (*params_copy)
87 string_list_split_in_place(¶ms, params_copy, ',', -1);
88 for (i = 0; i < params.nr; i++) {
89 const char *p = params.items[i].string;
90 if (!strcmp(p, "changes")) {
91 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
92 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
93 } else if (!strcmp(p, "lines")) {
94 DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
95 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
96 } else if (!strcmp(p, "files")) {
97 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
98 DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
99 } else if (!strcmp(p, "noncumulative")) {
100 DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
101 } else if (!strcmp(p, "cumulative")) {
102 DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
103 } else if (isdigit(*p)) {
104 char *end;
105 int permille = strtoul(p, &end, 10) * 10;
106 if (*end == '.' && isdigit(*++end)) {
107 /* only use first digit */
108 permille += *end - '0';
109 /* .. and ignore any further digits */
110 while (isdigit(*++end))
111 ; /* nothing */
112 }
113 if (!*end)
114 options->dirstat_permille = permille;
115 else {
116 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%s'\n"),
117 p);
118 ret++;
119 }
120 } else {
121 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%s'\n"), p);
122 ret++;
123 }
124 }
126 string_list_clear(¶ms, 0);
127 free(params_copy);
128 return ret;
129}
130static int parse_submodule_params(struct diff_options *options, const char *value)
132{
133 if (!strcmp(value, "log"))
134 DIFF_OPT_SET(options, SUBMODULE_LOG);
135 else if (!strcmp(value, "short"))
136 DIFF_OPT_CLR(options, SUBMODULE_LOG);
137 else
138 return -1;
139 return 0;
140}
141static int git_config_rename(const char *var, const char *value)
143{
144 if (!value)
145 return DIFF_DETECT_RENAME;
146 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
147 return DIFF_DETECT_COPY;
148 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
149}
150long parse_algorithm_value(const char *value)
152{
153 if (!value)
154 return -1;
155 else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
156 return 0;
157 else if (!strcasecmp(value, "minimal"))
158 return XDF_NEED_MINIMAL;
159 else if (!strcasecmp(value, "patience"))
160 return XDF_PATIENCE_DIFF;
161 else if (!strcasecmp(value, "histogram"))
162 return XDF_HISTOGRAM_DIFF;
163 return -1;
164}
165/*
167 * These are to give UI layer defaults.
168 * The core-level commands such as git-diff-files should
169 * never be affected by the setting of diff.renames
170 * the user happens to have in the configuration file.
171 */
172void init_diff_ui_defaults(void)
173{
174 diff_detect_rename_default = 1;
175}
176int git_diff_ui_config(const char *var, const char *value, void *cb)
178{
179 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
180 diff_use_color_default = git_config_colorbool(var, value);
181 return 0;
182 }
183 if (!strcmp(var, "diff.context")) {
184 diff_context_default = git_config_int(var, value);
185 if (diff_context_default < 0)
186 return -1;
187 return 0;
188 }
189 if (!strcmp(var, "diff.renames")) {
190 diff_detect_rename_default = git_config_rename(var, value);
191 return 0;
192 }
193 if (!strcmp(var, "diff.compactionheuristic")) {
194 diff_compaction_heuristic = git_config_bool(var, value);
195 return 0;
196 }
197 if (!strcmp(var, "diff.autorefreshindex")) {
198 diff_auto_refresh_index = git_config_bool(var, value);
199 return 0;
200 }
201 if (!strcmp(var, "diff.mnemonicprefix")) {
202 diff_mnemonic_prefix = git_config_bool(var, value);
203 return 0;
204 }
205 if (!strcmp(var, "diff.noprefix")) {
206 diff_no_prefix = git_config_bool(var, value);
207 return 0;
208 }
209 if (!strcmp(var, "diff.statgraphwidth")) {
210 diff_stat_graph_width = git_config_int(var, value);
211 return 0;
212 }
213 if (!strcmp(var, "diff.external"))
214 return git_config_string(&external_diff_cmd_cfg, var, value);
215 if (!strcmp(var, "diff.wordregex"))
216 return git_config_string(&diff_word_regex_cfg, var, value);
217 if (!strcmp(var, "diff.orderfile"))
218 return git_config_pathname(&diff_order_file_cfg, var, value);
219 if (!strcmp(var, "diff.ignoresubmodules"))
221 handle_ignore_submodules_arg(&default_diff_options, value);
222 if (!strcmp(var, "diff.submodule")) {
224 if (parse_submodule_params(&default_diff_options, value))
225 warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
226 value);
227 return 0;
228 }
229 if (!strcmp(var, "diff.algorithm")) {
231 diff_algorithm = parse_algorithm_value(value);
232 if (diff_algorithm < 0)
233 return -1;
234 return 0;
235 }
236 if (git_color_config(var, value, cb) < 0)
238 return -1;
239 return git_diff_basic_config(var, value, cb);
241}
242int git_diff_basic_config(const char *var, const char *value, void *cb)
244{
245 const char *name;
246 if (!strcmp(var, "diff.renamelimit")) {
248 diff_rename_limit_default = git_config_int(var, value);
249 return 0;
250 }
251 if (userdiff_config(var, value) < 0)
253 return -1;
254 if (skip_prefix(var, "diff.color.", &name) ||
256 skip_prefix(var, "color.diff.", &name)) {
257 int slot = parse_diff_color_slot(name);
258 if (slot < 0)
259 return 0;
260 if (!value)
261 return config_error_nonbool(var);
262 return color_parse(value, diff_colors[slot]);
263 }
264 /* like GNU diff's --suppress-blank-empty option */
266 if (!strcmp(var, "diff.suppressblankempty") ||
267 /* for backwards compatibility */
268 !strcmp(var, "diff.suppress-blank-empty")) {
269 diff_suppress_blank_empty = git_config_bool(var, value);
270 return 0;
271 }
272 if (!strcmp(var, "diff.dirstat")) {
274 struct strbuf errmsg = STRBUF_INIT;
275 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
276 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
277 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
278 errmsg.buf);
279 strbuf_release(&errmsg);
280 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
281 return 0;
282 }
283 if (starts_with(var, "submodule."))
285 return parse_submodule_config_option(var, value);
286 return git_default_config(var, value, cb);
288}
289static char *quote_two(const char *one, const char *two)
291{
292 int need_one = quote_c_style(one, NULL, NULL, 1);
293 int need_two = quote_c_style(two, NULL, NULL, 1);
294 struct strbuf res = STRBUF_INIT;
295 if (need_one + need_two) {
297 strbuf_addch(&res, '"');
298 quote_c_style(one, &res, NULL, 1);
299 quote_c_style(two, &res, NULL, 1);
300 strbuf_addch(&res, '"');
301 } else {
302 strbuf_addstr(&res, one);
303 strbuf_addstr(&res, two);
304 }
305 return strbuf_detach(&res, NULL);
306}
307static const char *external_diff(void)
309{
310 static const char *external_diff_cmd = NULL;
311 static int done_preparing = 0;
312 if (done_preparing)
314 return external_diff_cmd;
315 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
316 if (!external_diff_cmd)
317 external_diff_cmd = external_diff_cmd_cfg;
318 done_preparing = 1;
319 return external_diff_cmd;
320}
321/*
323 * Keep track of files used for diffing. Sometimes such an entry
324 * refers to a temporary file, sometimes to an existing file, and
325 * sometimes to "/dev/null".
326 */
327static struct diff_tempfile {
328 /*
329 * filename external diff should read from, or NULL if this
330 * entry is currently not in use:
331 */
332 const char *name;
333 char hex[GIT_SHA1_HEXSZ + 1];
335 char mode[10];
336 /*
338 * If this diff_tempfile instance refers to a temporary file,
339 * this tempfile object is used to manage its lifetime.
340 */
341 struct tempfile tempfile;
342} diff_temp[2];
343typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
345struct emit_callback {
347 int color_diff;
348 unsigned ws_rule;
349 int blank_at_eof_in_preimage;
350 int blank_at_eof_in_postimage;
351 int lno_in_preimage;
352 int lno_in_postimage;
353 sane_truncate_fn truncate;
354 const char **label_path;
355 struct diff_words_data *diff_words;
356 struct diff_options *opt;
357 struct strbuf *header;
358};
359static int count_lines(const char *data, int size)
361{
362 int count, ch, completely_empty = 1, nl_just_seen = 0;
363 count = 0;
364 while (0 < size--) {
365 ch = *data++;
366 if (ch == '\n') {
367 count++;
368 nl_just_seen = 1;
369 completely_empty = 0;
370 }
371 else {
372 nl_just_seen = 0;
373 completely_empty = 0;
374 }
375 }
376 if (completely_empty)
377 return 0;
378 if (!nl_just_seen)
379 count++; /* no trailing newline */
380 return count;
381}
382static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
384{
385 if (!DIFF_FILE_VALID(one)) {
386 mf->ptr = (char *)""; /* does not matter */
387 mf->size = 0;
388 return 0;
389 }
390 else if (diff_populate_filespec(one, 0))
391 return -1;
392 mf->ptr = one->data;
394 mf->size = one->size;
395 return 0;
396}
397/* like fill_mmfile, but only for size, so we can avoid retrieving blob */
399static unsigned long diff_filespec_size(struct diff_filespec *one)
400{
401 if (!DIFF_FILE_VALID(one))
402 return 0;
403 diff_populate_filespec(one, CHECK_SIZE_ONLY);
404 return one->size;
405}
406static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
408{
409 char *ptr = mf->ptr;
410 long size = mf->size;
411 int cnt = 0;
412 if (!size)
414 return cnt;
415 ptr += size - 1; /* pointing at the very end */
416 if (*ptr != '\n')
417 ; /* incomplete line */
418 else
419 ptr--; /* skip the last LF */
420 while (mf->ptr < ptr) {
421 char *prev_eol;
422 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
423 if (*prev_eol == '\n')
424 break;
425 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
426 break;
427 cnt++;
428 ptr = prev_eol - 1;
429 }
430 return cnt;
431}
432static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
434 struct emit_callback *ecbdata)
435{
436 int l1, l2, at;
437 unsigned ws_rule = ecbdata->ws_rule;
438 l1 = count_trailing_blank(mf1, ws_rule);
439 l2 = count_trailing_blank(mf2, ws_rule);
440 if (l2 <= l1) {
441 ecbdata->blank_at_eof_in_preimage = 0;
442 ecbdata->blank_at_eof_in_postimage = 0;
443 return;
444 }
445 at = count_lines(mf1->ptr, mf1->size);
446 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
447 at = count_lines(mf2->ptr, mf2->size);
449 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
450}
451static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
453 int first, const char *line, int len)
454{
455 int has_trailing_newline, has_trailing_carriage_return;
456 int nofirst;
457 FILE *file = o->file;
458 fputs(diff_line_prefix(o), file);
460 if (len == 0) {
462 has_trailing_newline = (first == '\n');
463 has_trailing_carriage_return = (!has_trailing_newline &&
464 (first == '\r'));
465 nofirst = has_trailing_newline || has_trailing_carriage_return;
466 } else {
467 has_trailing_newline = (len > 0 && line[len-1] == '\n');
468 if (has_trailing_newline)
469 len--;
470 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
471 if (has_trailing_carriage_return)
472 len--;
473 nofirst = 0;
474 }
475 if (len || !nofirst) {
477 fputs(set, file);
478 if (!nofirst)
479 fputc(first, file);
480 fwrite(line, len, 1, file);
481 fputs(reset, file);
482 }
483 if (has_trailing_carriage_return)
484 fputc('\r', file);
485 if (has_trailing_newline)
486 fputc('\n', file);
487}
488static void emit_line(struct diff_options *o, const char *set, const char *reset,
490 const char *line, int len)
491{
492 emit_line_0(o, set, reset, line[0], line+1, len-1);
493}
494static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
496{
497 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
498 ecbdata->blank_at_eof_in_preimage &&
499 ecbdata->blank_at_eof_in_postimage &&
500 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
501 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
502 return 0;
503 return ws_blank_line(line, len, ecbdata->ws_rule);
504}
505static void emit_line_checked(const char *reset,
507 struct emit_callback *ecbdata,
508 const char *line, int len,
509 enum color_diff color,
510 unsigned ws_error_highlight,
511 char sign)
512{
513 const char *set = diff_get_color(ecbdata->color_diff, color);
514 const char *ws = NULL;
515 if (ecbdata->opt->ws_error_highlight & ws_error_highlight) {
517 ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
518 if (!*ws)
519 ws = NULL;
520 }
521 if (!ws)
523 emit_line_0(ecbdata->opt, set, reset, sign, line, len);
524 else if (sign == '+' && new_blank_line_at_eof(ecbdata, line, len))
525 /* Blank line at EOF - paint '+' as well */
526 emit_line_0(ecbdata->opt, ws, reset, sign, line, len);
527 else {
528 /* Emit just the prefix, then the rest. */
529 emit_line_0(ecbdata->opt, set, reset, sign, "", 0);
530 ws_check_emit(line, len, ecbdata->ws_rule,
531 ecbdata->opt->file, set, reset, ws);
532 }
533}
534static void emit_add_line(const char *reset,
536 struct emit_callback *ecbdata,
537 const char *line, int len)
538{
539 emit_line_checked(reset, ecbdata, line, len,
540 DIFF_FILE_NEW, WSEH_NEW, '+');
541}
542static void emit_del_line(const char *reset,
544 struct emit_callback *ecbdata,
545 const char *line, int len)
546{
547 emit_line_checked(reset, ecbdata, line, len,
548 DIFF_FILE_OLD, WSEH_OLD, '-');
549}
550static void emit_context_line(const char *reset,
552 struct emit_callback *ecbdata,
553 const char *line, int len)
554{
555 emit_line_checked(reset, ecbdata, line, len,
556 DIFF_CONTEXT, WSEH_CONTEXT, ' ');
557}
558static void emit_hunk_header(struct emit_callback *ecbdata,
560 const char *line, int len)
561{
562 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
563 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
564 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
565 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
566 static const char atat[2] = { '@', '@' };
567 const char *cp, *ep;
568 struct strbuf msgbuf = STRBUF_INIT;
569 int org_len = len;
570 int i = 1;
571 /*
573 * As a hunk header must begin with "@@ -<old>, +<new> @@",
574 * it always is at least 10 bytes long.
575 */
576 if (len < 10 ||
577 memcmp(line, atat, 2) ||
578 !(ep = memmem(line + 2, len - 2, atat, 2))) {
579 emit_line(ecbdata->opt, context, reset, line, len);
580 return;
581 }
582 ep += 2; /* skip over @@ */
583 /* The hunk header in fraginfo color */
585 strbuf_addstr(&msgbuf, frag);
586 strbuf_add(&msgbuf, line, ep - line);
587 strbuf_addstr(&msgbuf, reset);
588 /*
590 * trailing "\r\n"
591 */
592 for ( ; i < 3; i++)
593 if (line[len - i] == '\r' || line[len - i] == '\n')
594 len--;
595 /* blank before the func header */
597 for (cp = ep; ep - line < len; ep++)
598 if (*ep != ' ' && *ep != '\t')
599 break;
600 if (ep != cp) {
601 strbuf_addstr(&msgbuf, context);
602 strbuf_add(&msgbuf, cp, ep - cp);
603 strbuf_addstr(&msgbuf, reset);
604 }
605 if (ep < line + len) {
607 strbuf_addstr(&msgbuf, func);
608 strbuf_add(&msgbuf, ep, line + len - ep);
609 strbuf_addstr(&msgbuf, reset);
610 }
611 strbuf_add(&msgbuf, line + len, org_len - len);
613 emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
614 strbuf_release(&msgbuf);
615}
616static struct diff_tempfile *claim_diff_tempfile(void) {
618 int i;
619 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
620 if (!diff_temp[i].name)
621 return diff_temp + i;
622 die("BUG: diff is failing to clean up its tempfiles");
623}
624static void remove_tempfile(void)
626{
627 int i;
628 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
629 if (is_tempfile_active(&diff_temp[i].tempfile))
630 delete_tempfile(&diff_temp[i].tempfile);
631 diff_temp[i].name = NULL;
632 }
633}
634static void print_line_count(FILE *file, int count)
636{
637 switch (count) {
638 case 0:
639 fprintf(file, "0,0");
640 break;
641 case 1:
642 fprintf(file, "1");
643 break;
644 default:
645 fprintf(file, "1,%d", count);
646 break;
647 }
648}
649static void emit_rewrite_lines(struct emit_callback *ecb,
651 int prefix, const char *data, int size)
652{
653 const char *endp = NULL;
654 static const char *nneof = " No newline at end of file\n";
655 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
656 while (0 < size) {
658 int len;
659 endp = memchr(data, '\n', size);
661 len = endp ? (endp - data + 1) : size;
662 if (prefix != '+') {
663 ecb->lno_in_preimage++;
664 emit_del_line(reset, ecb, data, len);
665 } else {
666 ecb->lno_in_postimage++;
667 emit_add_line(reset, ecb, data, len);
668 }
669 size -= len;
670 data += len;
671 }
672 if (!endp) {
673 const char *context = diff_get_color(ecb->color_diff,
674 DIFF_CONTEXT);
675 putc('\n', ecb->opt->file);
676 emit_line_0(ecb->opt, context, reset, '\\',
677 nneof, strlen(nneof));
678 }
679}
680static void emit_rewrite_diff(const char *name_a,
682 const char *name_b,
683 struct diff_filespec *one,
684 struct diff_filespec *two,
685 struct userdiff_driver *textconv_one,
686 struct userdiff_driver *textconv_two,
687 struct diff_options *o)
688{
689 int lc_a, lc_b;
690 const char *name_a_tab, *name_b_tab;
691 const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
692 const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
693 const char *reset = diff_get_color(o->use_color, DIFF_RESET);
694 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
695 const char *a_prefix, *b_prefix;
696 char *data_one, *data_two;
697 size_t size_one, size_two;
698 struct emit_callback ecbdata;
699 const char *line_prefix = diff_line_prefix(o);
700 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
702 a_prefix = o->b_prefix;
703 b_prefix = o->a_prefix;
704 } else {
705 a_prefix = o->a_prefix;
706 b_prefix = o->b_prefix;
707 }
708 name_a += (*name_a == '/');
710 name_b += (*name_b == '/');
711 name_a_tab = strchr(name_a, ' ') ? "\t" : "";
712 name_b_tab = strchr(name_b, ' ') ? "\t" : "";
713 strbuf_reset(&a_name);
715 strbuf_reset(&b_name);
716 quote_two_c_style(&a_name, a_prefix, name_a, 0);
717 quote_two_c_style(&b_name, b_prefix, name_b, 0);
718 size_one = fill_textconv(textconv_one, one, &data_one);
720 size_two = fill_textconv(textconv_two, two, &data_two);
721 memset(&ecbdata, 0, sizeof(ecbdata));
723 ecbdata.color_diff = want_color(o->use_color);
724 ecbdata.ws_rule = whitespace_rule(name_b);
725 ecbdata.opt = o;
726 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
727 mmfile_t mf1, mf2;
728 mf1.ptr = (char *)data_one;
729 mf2.ptr = (char *)data_two;
730 mf1.size = size_one;
731 mf2.size = size_two;
732 check_blank_at_eof(&mf1, &mf2, &ecbdata);
733 }
734 ecbdata.lno_in_preimage = 1;
735 ecbdata.lno_in_postimage = 1;
736 lc_a = count_lines(data_one, size_one);
738 lc_b = count_lines(data_two, size_two);
739 fprintf(o->file,
740 "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
741 line_prefix, metainfo, a_name.buf, name_a_tab, reset,
742 line_prefix, metainfo, b_name.buf, name_b_tab, reset,
743 line_prefix, fraginfo);
744 if (!o->irreversible_delete)
745 print_line_count(o->file, lc_a);
746 else
747 fprintf(o->file, "?,?");
748 fprintf(o->file, " +");
749 print_line_count(o->file, lc_b);
750 fprintf(o->file, " @@%s\n", reset);
751 if (lc_a && !o->irreversible_delete)
752 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
753 if (lc_b)
754 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
755 if (textconv_one)
756 free((char *)data_one);
757 if (textconv_two)
758 free((char *)data_two);
759}
760struct diff_words_buffer {
762 mmfile_t text;
763 long alloc;
764 struct diff_words_orig {
765 const char *begin, *end;
766 } *orig;
767 int orig_nr, orig_alloc;
768};
769static void diff_words_append(char *line, unsigned long len,
771 struct diff_words_buffer *buffer)
772{
773 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
774 line++;
775 len--;
776 memcpy(buffer->text.ptr + buffer->text.size, line, len);
777 buffer->text.size += len;
778 buffer->text.ptr[buffer->text.size] = '\0';
779}
780struct diff_words_style_elem {
782 const char *prefix;
783 const char *suffix;
784 const char *color; /* NULL; filled in by the setup code if
785 * color is enabled */
786};
787struct diff_words_style {
789 enum diff_words_type type;
790 struct diff_words_style_elem new, old, ctx;
791 const char *newline;
792};
793static struct diff_words_style diff_words_styles[] = {
795 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
796 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
797 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
798};
799struct diff_words_data {
801 struct diff_words_buffer minus, plus;
802 const char *current_plus;
803 int last_minus;
804 struct diff_options *opt;
805 regex_t *word_regex;
806 enum diff_words_type type;
807 struct diff_words_style *style;
808};
809static int fn_out_diff_words_write_helper(FILE *fp,
811 struct diff_words_style_elem *st_el,
812 const char *newline,
813 size_t count, const char *buf,
814 const char *line_prefix)
815{
816 int print = 0;
817 while (count) {
819 char *p = memchr(buf, '\n', count);
820 if (print)
821 fputs(line_prefix, fp);
822 if (p != buf) {
823 if (st_el->color && fputs(st_el->color, fp) < 0)
824 return -1;
825 if (fputs(st_el->prefix, fp) < 0 ||
826 fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
827 fputs(st_el->suffix, fp) < 0)
828 return -1;
829 if (st_el->color && *st_el->color
830 && fputs(GIT_COLOR_RESET, fp) < 0)
831 return -1;
832 }
833 if (!p)
834 return 0;
835 if (fputs(newline, fp) < 0)
836 return -1;
837 count -= p + 1 - buf;
838 buf = p + 1;
839 print = 1;
840 }
841 return 0;
842}
843/*
845 * '--color-words' algorithm can be described as:
846 *
847 * 1. collect a the minus/plus lines of a diff hunk, divided into
848 * minus-lines and plus-lines;
849 *
850 * 2. break both minus-lines and plus-lines into words and
851 * place them into two mmfile_t with one word for each line;
852 *
853 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
854 *
855 * And for the common parts of the both file, we output the plus side text.
856 * diff_words->current_plus is used to trace the current position of the plus file
857 * which printed. diff_words->last_minus is used to trace the last minus word
858 * printed.
859 *
860 * For '--graph' to work with '--color-words', we need to output the graph prefix
861 * on each line of color words output. Generally, there are two conditions on
862 * which we should output the prefix.
863 *
864 * 1. diff_words->last_minus == 0 &&
865 * diff_words->current_plus == diff_words->plus.text.ptr
866 *
867 * that is: the plus text must start as a new line, and if there is no minus
868 * word printed, a graph prefix must be printed.
869 *
870 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
871 * *(diff_words->current_plus - 1) == '\n'
872 *
873 * that is: a graph prefix must be printed following a '\n'
874 */
875static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
876{
877 if ((diff_words->last_minus == 0 &&
878 diff_words->current_plus == diff_words->plus.text.ptr) ||
879 (diff_words->current_plus > diff_words->plus.text.ptr &&
880 *(diff_words->current_plus - 1) == '\n')) {
881 return 1;
882 } else {
883 return 0;
884 }
885}
886static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
888{
889 struct diff_words_data *diff_words = priv;
890 struct diff_words_style *style = diff_words->style;
891 int minus_first, minus_len, plus_first, plus_len;
892 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
893 struct diff_options *opt = diff_words->opt;
894 const char *line_prefix;
895 if (line[0] != '@' || parse_hunk_header(line, len,
897 &minus_first, &minus_len, &plus_first, &plus_len))
898 return;
899 assert(opt);
901 line_prefix = diff_line_prefix(opt);
902 /* POSIX requires that first be decremented by one if len == 0... */
904 if (minus_len) {
905 minus_begin = diff_words->minus.orig[minus_first].begin;
906 minus_end =
907 diff_words->minus.orig[minus_first + minus_len - 1].end;
908 } else
909 minus_begin = minus_end =
910 diff_words->minus.orig[minus_first].end;
911 if (plus_len) {
913 plus_begin = diff_words->plus.orig[plus_first].begin;
914 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
915 } else
916 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
917 if (color_words_output_graph_prefix(diff_words)) {
919 fputs(line_prefix, diff_words->opt->file);
920 }
921 if (diff_words->current_plus != plus_begin) {
922 fn_out_diff_words_write_helper(diff_words->opt->file,
923 &style->ctx, style->newline,
924 plus_begin - diff_words->current_plus,
925 diff_words->current_plus, line_prefix);
926 if (*(plus_begin - 1) == '\n')
927 fputs(line_prefix, diff_words->opt->file);
928 }
929 if (minus_begin != minus_end) {
930 fn_out_diff_words_write_helper(diff_words->opt->file,
931 &style->old, style->newline,
932 minus_end - minus_begin, minus_begin,
933 line_prefix);
934 }
935 if (plus_begin != plus_end) {
936 fn_out_diff_words_write_helper(diff_words->opt->file,
937 &style->new, style->newline,
938 plus_end - plus_begin, plus_begin,
939 line_prefix);
940 }
941 diff_words->current_plus = plus_end;
943 diff_words->last_minus = minus_first;
944}
945/* This function starts looking at *begin, and returns 0 iff a word was found. */
947static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
948 int *begin, int *end)
949{
950 if (word_regex && *begin < buffer->size) {
951 regmatch_t match[1];
952 if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
953 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
954 '\n', match[0].rm_eo - match[0].rm_so);
955 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
956 *begin += match[0].rm_so;
957 return *begin >= *end;
958 }
959 return -1;
960 }
961 /* find the next word */
963 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
964 (*begin)++;
965 if (*begin >= buffer->size)
966 return -1;
967 /* find the end of the word */
969 *end = *begin + 1;
970 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
971 (*end)++;
972 return 0;
974}
975/*
977 * This function splits the words in buffer->text, stores the list with
978 * newline separator into out, and saves the offsets of the original words
979 * in buffer->orig.
980 */
981static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
982 regex_t *word_regex)
983{
984 int i, j;
985 long alloc = 0;
986 out->size = 0;
988 out->ptr = NULL;
989 /* fake an empty "0th" word */
991 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
992 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
993 buffer->orig_nr = 1;
994 for (i = 0; i < buffer->text.size; i++) {
996 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
997 return;
998 /* store original boundaries */
1000 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
1001 buffer->orig_alloc);
1002 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
1003 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
1004 buffer->orig_nr++;
1005 /* store one word */
1007 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
1008 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
1009 out->ptr[out->size + j - i] = '\n';
1010 out->size += j - i + 1;
1011 i = j - 1;
1013 }
1014}
1015/* this executes the word diff on the accumulated buffers */
1017static void diff_words_show(struct diff_words_data *diff_words)
1018{
1019 xpparam_t xpp;
1020 xdemitconf_t xecfg;
1021 mmfile_t minus, plus;
1022 struct diff_words_style *style = diff_words->style;
1023 struct diff_options *opt = diff_words->opt;
1025 const char *line_prefix;
1026 assert(opt);
1028 line_prefix = diff_line_prefix(opt);
1029 /* special case: only removal */
1031 if (!diff_words->plus.text.size) {
1032 fputs(line_prefix, diff_words->opt->file);
1033 fn_out_diff_words_write_helper(diff_words->opt->file,
1034 &style->old, style->newline,
1035 diff_words->minus.text.size,
1036 diff_words->minus.text.ptr, line_prefix);
1037 diff_words->minus.text.size = 0;
1038 return;
1039 }
1040 diff_words->current_plus = diff_words->plus.text.ptr;
1042 diff_words->last_minus = 0;
1043 memset(&xpp, 0, sizeof(xpp));
1045 memset(&xecfg, 0, sizeof(xecfg));
1046 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1047 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1048 xpp.flags = 0;
1049 /* as only the hunk header will be parsed, we need a 0-context */
1050 xecfg.ctxlen = 0;
1051 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1052 &xpp, &xecfg))
1053 die("unable to generate word diff");
1054 free(minus.ptr);
1055 free(plus.ptr);
1056 if (diff_words->current_plus != diff_words->plus.text.ptr +
1057 diff_words->plus.text.size) {
1058 if (color_words_output_graph_prefix(diff_words))
1059 fputs(line_prefix, diff_words->opt->file);
1060 fn_out_diff_words_write_helper(diff_words->opt->file,
1061 &style->ctx, style->newline,
1062 diff_words->plus.text.ptr + diff_words->plus.text.size
1063 - diff_words->current_plus, diff_words->current_plus,
1064 line_prefix);
1065 }
1066 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1067}
1068/* In "color-words" mode, show word-diff of words accumulated in the buffer */
1070static void diff_words_flush(struct emit_callback *ecbdata)
1071{
1072 if (ecbdata->diff_words->minus.text.size ||
1073 ecbdata->diff_words->plus.text.size)
1074 diff_words_show(ecbdata->diff_words);
1075}
1076static void diff_filespec_load_driver(struct diff_filespec *one)
1078{
1079 /* Use already-loaded driver */
1080 if (one->driver)
1081 return;
1082 if (S_ISREG(one->mode))
1084 one->driver = userdiff_find_by_path(one->path);
1085 /* Fallback to default settings */
1087 if (!one->driver)
1088 one->driver = userdiff_find_by_name("default");
1089}
1090static const char *userdiff_word_regex(struct diff_filespec *one)
1092{
1093 diff_filespec_load_driver(one);
1094 return one->driver->word_regex;
1095}
1096static void init_diff_words_data(struct emit_callback *ecbdata,
1098 struct diff_options *orig_opts,
1099 struct diff_filespec *one,
1100 struct diff_filespec *two)
1101{
1102 int i;
1103 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1104 memcpy(o, orig_opts, sizeof(struct diff_options));
1105 ecbdata->diff_words =
1107 xcalloc(1, sizeof(struct diff_words_data));
1108 ecbdata->diff_words->type = o->word_diff;
1109 ecbdata->diff_words->opt = o;
1110 if (!o->word_regex)
1111 o->word_regex = userdiff_word_regex(one);
1112 if (!o->word_regex)
1113 o->word_regex = userdiff_word_regex(two);
1114 if (!o->word_regex)
1115 o->word_regex = diff_word_regex_cfg;
1116 if (o->word_regex) {
1117 ecbdata->diff_words->word_regex = (regex_t *)
1118 xmalloc(sizeof(regex_t));
1119 if (regcomp(ecbdata->diff_words->word_regex,
1120 o->word_regex,
1121 REG_EXTENDED | REG_NEWLINE))
1122 die ("Invalid regular expression: %s",
1123 o->word_regex);
1124 }
1125 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1126 if (o->word_diff == diff_words_styles[i].type) {
1127 ecbdata->diff_words->style =
1128 &diff_words_styles[i];
1129 break;
1130 }
1131 }
1132 if (want_color(o->use_color)) {
1133 struct diff_words_style *st = ecbdata->diff_words->style;
1134 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1135 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1136 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
1137 }
1138}
1139static void free_diff_words_data(struct emit_callback *ecbdata)
1141{
1142 if (ecbdata->diff_words) {
1143 diff_words_flush(ecbdata);
1144 free (ecbdata->diff_words->opt);
1145 free (ecbdata->diff_words->minus.text.ptr);
1146 free (ecbdata->diff_words->minus.orig);
1147 free (ecbdata->diff_words->plus.text.ptr);
1148 free (ecbdata->diff_words->plus.orig);
1149 if (ecbdata->diff_words->word_regex) {
1150 regfree(ecbdata->diff_words->word_regex);
1151 free(ecbdata->diff_words->word_regex);
1152 }
1153 free(ecbdata->diff_words);
1154 ecbdata->diff_words = NULL;
1155 }
1156}
1157const char *diff_get_color(int diff_use_color, enum color_diff ix)
1159{
1160 if (want_color(diff_use_color))
1161 return diff_colors[ix];
1162 return "";
1163}
1164const char *diff_line_prefix(struct diff_options *opt)
1166{
1167 struct strbuf *msgbuf;
1168 if (!opt->output_prefix)
1169 return "";
1170 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1172 return msgbuf->buf;
1173}
1174static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1176{
1177 const char *cp;
1178 unsigned long allot;
1179 size_t l = len;
1180 if (ecb->truncate)
1182 return ecb->truncate(line, len);
1183 cp = line;
1184 allot = l;
1185 while (0 < l) {
1186 (void) utf8_width(&cp, &l);
1187 if (!cp)
1188 break; /* truncated in the middle? */
1189 }
1190 return allot - l;
1191}
1192static void find_lno(const char *line, struct emit_callback *ecbdata)
1194{
1195 const char *p;
1196 ecbdata->lno_in_preimage = 0;
1197 ecbdata->lno_in_postimage = 0;
1198 p = strchr(line, '-');
1199 if (!p)
1200 return; /* cannot happen */
1201 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1202 p = strchr(p, '+');
1203 if (!p)
1204 return; /* cannot happen */
1205 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1206}
1207static void fn_out_consume(void *priv, char *line, unsigned long len)
1209{
1210 struct emit_callback *ecbdata = priv;
1211 const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1212 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
1213 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1214 struct diff_options *o = ecbdata->opt;
1215 const char *line_prefix = diff_line_prefix(o);
1216 o->found_changes = 1;
1218 if (ecbdata->header) {
1220 fprintf(o->file, "%s", ecbdata->header->buf);
1221 strbuf_reset(ecbdata->header);
1222 ecbdata->header = NULL;
1223 }
1224 if (ecbdata->label_path[0]) {
1226 const char *name_a_tab, *name_b_tab;
1227 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1229 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1230 fprintf(o->file, "%s%s--- %s%s%s\n",
1232 line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1233 fprintf(o->file, "%s%s+++ %s%s%s\n",
1234 line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1235 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1236 }
1237 if (diff_suppress_blank_empty
1239 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1240 line[0] = '\n';
1241 len = 1;
1242 }
1243 if (line[0] == '@') {
1245 if (ecbdata->diff_words)
1246 diff_words_flush(ecbdata);
1247 len = sane_truncate_line(ecbdata, line, len);
1248 find_lno(line, ecbdata);
1249 emit_hunk_header(ecbdata, line, len);
1250 if (line[len-1] != '\n')
1251 putc('\n', o->file);
1252 return;
1253 }
1254 if (len < 1) {
1256 emit_line(o, reset, reset, line, len);
1257 if (ecbdata->diff_words
1258 && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1259 fputs("~\n", o->file);
1260 return;
1261 }
1262 if (ecbdata->diff_words) {
1264 if (line[0] == '-') {
1265 diff_words_append(line, len,
1266 &ecbdata->diff_words->minus);
1267 return;
1268 } else if (line[0] == '+') {
1269 diff_words_append(line, len,
1270 &ecbdata->diff_words->plus);
1271 return;
1272 } else if (starts_with(line, "\\ ")) {
1273 /*
1274 * Eat the "no newline at eof" marker as if we
1275 * saw a "+" or "-" line with nothing on it,
1276 * and return without diff_words_flush() to
1277 * defer processing. If this is the end of
1278 * preimage, more "+" lines may come after it.
1279 */
1280 return;
1281 }
1282 diff_words_flush(ecbdata);
1283 if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1284 emit_line(o, context, reset, line, len);
1285 fputs("~\n", o->file);
1286 } else {
1287 /*
1288 * Skip the prefix character, if any. With
1289 * diff_suppress_blank_empty, there may be
1290 * none.
1291 */
1292 if (line[0] != '\n') {
1293 line++;
1294 len--;
1295 }
1296 emit_line(o, context, reset, line, len);
1297 }
1298 return;
1299 }
1300 switch (line[0]) {
1302 case '+':
1303 ecbdata->lno_in_postimage++;
1304 emit_add_line(reset, ecbdata, line + 1, len - 1);
1305 break;
1306 case '-':
1307 ecbdata->lno_in_preimage++;
1308 emit_del_line(reset, ecbdata, line + 1, len - 1);
1309 break;
1310 case ' ':
1311 ecbdata->lno_in_postimage++;
1312 ecbdata->lno_in_preimage++;
1313 emit_context_line(reset, ecbdata, line + 1, len - 1);
1314 break;
1315 default:
1316 /* incomplete line at the end */
1317 ecbdata->lno_in_preimage++;
1318 emit_line(o, diff_get_color(ecbdata->color_diff, DIFF_CONTEXT),
1319 reset, line, len);
1320 break;
1321 }
1322}
1323static char *pprint_rename(const char *a, const char *b)
1325{
1326 const char *old = a;
1327 const char *new = b;
1328 struct strbuf name = STRBUF_INIT;
1329 int pfx_length, sfx_length;
1330 int pfx_adjust_for_slash;
1331 int len_a = strlen(a);
1332 int len_b = strlen(b);
1333 int a_midlen, b_midlen;
1334 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1335 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1336 if (qlen_a || qlen_b) {
1338 quote_c_style(a, &name, NULL, 0);
1339 strbuf_addstr(&name, " => ");
1340 quote_c_style(b, &name, NULL, 0);
1341 return strbuf_detach(&name, NULL);
1342 }
1343 /* Find common prefix */
1345 pfx_length = 0;
1346 while (*old && *new && *old == *new) {
1347 if (*old == '/')
1348 pfx_length = old - a + 1;
1349 old++;
1350 new++;
1351 }
1352 /* Find common suffix */
1354 old = a + len_a;
1355 new = b + len_b;
1356 sfx_length = 0;
1357 /*
1358 * If there is a common prefix, it must end in a slash. In
1359 * that case we let this loop run 1 into the prefix to see the
1360 * same slash.
1361 *
1362 * If there is no common prefix, we cannot do this as it would
1363 * underrun the input strings.
1364 */
1365 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
1366 while (a + pfx_length - pfx_adjust_for_slash <= old &&
1367 b + pfx_length - pfx_adjust_for_slash <= new &&
1368 *old == *new) {
1369 if (*old == '/')
1370 sfx_length = len_a - (old - a);
1371 old--;
1372 new--;
1373 }
1374 /*
1376 * pfx{mid-a => mid-b}sfx
1377 * {pfx-a => pfx-b}sfx
1378 * pfx{sfx-a => sfx-b}
1379 * name-a => name-b
1380 */
1381 a_midlen = len_a - pfx_length - sfx_length;
1382 b_midlen = len_b - pfx_length - sfx_length;
1383 if (a_midlen < 0)
1384 a_midlen = 0;
1385 if (b_midlen < 0)
1386 b_midlen = 0;
1387 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1389 if (pfx_length + sfx_length) {
1390 strbuf_add(&name, a, pfx_length);
1391 strbuf_addch(&name, '{');
1392 }
1393 strbuf_add(&name, a + pfx_length, a_midlen);
1394 strbuf_addstr(&name, " => ");
1395 strbuf_add(&name, b + pfx_length, b_midlen);
1396 if (pfx_length + sfx_length) {
1397 strbuf_addch(&name, '}');
1398 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1399 }
1400 return strbuf_detach(&name, NULL);
1401}
1402struct diffstat_t {
1404 int nr;
1405 int alloc;
1406 struct diffstat_file {
1407 char *from_name;
1408 char *name;
1409 char *print_name;
1410 unsigned is_unmerged:1;
1411 unsigned is_binary:1;
1412 unsigned is_renamed:1;
1413 unsigned is_interesting:1;
1414 uintmax_t added, deleted;
1415 } **files;
1416};
1417static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1419 const char *name_a,
1420 const char *name_b)
1421{
1422 struct diffstat_file *x;
1423 x = xcalloc(1, sizeof(*x));
1424 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
1425 diffstat->files[diffstat->nr++] = x;
1426 if (name_b) {
1427 x->from_name = xstrdup(name_a);
1428 x->name = xstrdup(name_b);
1429 x->is_renamed = 1;
1430 }
1431 else {
1432 x->from_name = NULL;
1433 x->name = xstrdup(name_a);
1434 }
1435 return x;
1436}
1437static void diffstat_consume(void *priv, char *line, unsigned long len)
1439{
1440 struct diffstat_t *diffstat = priv;
1441 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1442 if (line[0] == '+')
1444 x->added++;
1445 else if (line[0] == '-')
1446 x->deleted++;
1447}
1448const char mime_boundary_leader[] = "------------";
1450static int scale_linear(int it, int width, int max_change)
1452{
1453 if (!it)
1454 return 0;
1455 /*
1456 * make sure that at least one '-' or '+' is printed if
1457 * there is any change to this path. The easiest way is to
1458 * scale linearly as if the alloted width is one column shorter
1459 * than it is, and then add 1 to the result.
1460 */
1461 return 1 + (it * (width - 1) / max_change);
1462}
1463static void show_name(FILE *file,
1465 const char *prefix, const char *name, int len)
1466{
1467 fprintf(file, " %s%-*s |", prefix, len, name);
1468}
1469static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1471{
1472 if (cnt <= 0)
1473 return;
1474 fprintf(file, "%s", set);
1475 while (cnt--)
1476 putc(ch, file);
1477 fprintf(file, "%s", reset);
1478}
1479static void fill_print_name(struct diffstat_file *file)
1481{
1482 char *pname;
1483 if (file->print_name)
1485 return;
1486 if (!file->is_renamed) {
1488 struct strbuf buf = STRBUF_INIT;
1489 if (quote_c_style(file->name, &buf, NULL, 0)) {
1490 pname = strbuf_detach(&buf, NULL);
1491 } else {
1492 pname = file->name;
1493 strbuf_release(&buf);
1494 }
1495 } else {
1496 pname = pprint_rename(file->from_name, file->name);
1497 }
1498 file->print_name = pname;
1499}
1500int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1502{
1503 struct strbuf sb = STRBUF_INIT;
1504 int ret;
1505 if (!files) {
1507 assert(insertions == 0 && deletions == 0);
1508 return fprintf(fp, "%s\n", " 0 files changed");
1509 }
1510 strbuf_addf(&sb,
1512 (files == 1) ? " %d file changed" : " %d files changed",
1513 files);
1514 /*
1516 * For binary diff, the caller may want to print "x files
1517 * changed" with insertions == 0 && deletions == 0.
1518 *
1519 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1520 * is probably less confusing (i.e skip over "2 files changed
1521 * but nothing about added/removed lines? Is this a bug in Git?").
1522 */
1523 if (insertions || deletions == 0) {
1524 strbuf_addf(&sb,
1525 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1526 insertions);
1527 }
1528 if (deletions || insertions == 0) {
1530 strbuf_addf(&sb,
1531 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1532 deletions);
1533 }
1534 strbuf_addch(&sb, '\n');
1535 ret = fputs(sb.buf, fp);
1536 strbuf_release(&sb);
1537 return ret;
1538}
1539static void show_stats(struct diffstat_t *data, struct diff_options *options)
1541{
1542 int i, len, add, del, adds = 0, dels = 0;
1543 uintmax_t max_change = 0, max_len = 0;
1544 int total_files = data->nr, count;
1545 int width, name_width, graph_width, number_width = 0, bin_width = 0;
1546 const char *reset, *add_c, *del_c;
1547 const char *line_prefix = "";
1548 int extra_shown = 0;
1549 if (data->nr == 0)
1551 return;
1552 line_prefix = diff_line_prefix(options);
1554 count = options->stat_count ? options->stat_count : data->nr;
1555 reset = diff_get_color_opt(options, DIFF_RESET);
1557 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1558 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1559 /*
1561 * Find the longest filename and max number of changes
1562 */
1563 for (i = 0; (i < count) && (i < data->nr); i++) {
1564 struct diffstat_file *file = data->files[i];
1565 uintmax_t change = file->added + file->deleted;
1566 if (!file->is_interesting && (change == 0)) {
1568 count++; /* not shown == room for one more */
1569 continue;
1570 }
1571 fill_print_name(file);
1572 len = strlen(file->print_name);
1573 if (max_len < len)
1574 max_len = len;
1575 if (file->is_unmerged) {
1577 /* "Unmerged" is 8 characters */
1578 bin_width = bin_width < 8 ? 8 : bin_width;
1579 continue;
1580 }
1581 if (file->is_binary) {
1582 /* "Bin XXX -> YYY bytes" */
1583 int w = 14 + decimal_width(file->added)
1584 + decimal_width(file->deleted);
1585 bin_width = bin_width < w ? w : bin_width;
1586 /* Display change counts aligned with "Bin" */
1587 number_width = 3;
1588 continue;
1589 }
1590 if (max_change < change)
1592 max_change = change;
1593 }
1594 count = i; /* where we can stop scanning in data->files[] */
1595 /*
1597 * We have width = stat_width or term_columns() columns total.
1598 * We want a maximum of min(max_len, stat_name_width) for the name part.
1599 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1600 * We also need 1 for " " and 4 + decimal_width(max_change)
1601 * for " | NNNN " and one the empty column at the end, altogether
1602 * 6 + decimal_width(max_change).
1603 *
1604 * If there's not enough space, we will use the smaller of
1605 * stat_name_width (if set) and 5/8*width for the filename,
1606 * and the rest for constant elements + graph part, but no more
1607 * than stat_graph_width for the graph part.
1608 * (5/8 gives 50 for filename and 30 for the constant parts + graph
1609 * for the standard terminal size).
1610 *
1611 * In other words: stat_width limits the maximum width, and
1612 * stat_name_width fixes the maximum width of the filename,
1613 * and is also used to divide available columns if there
1614 * aren't enough.
1615 *
1616 * Binary files are displayed with "Bin XXX -> YYY bytes"
1617 * instead of the change count and graph. This part is treated
1618 * similarly to the graph part, except that it is not
1619 * "scaled". If total width is too small to accommodate the
1620 * guaranteed minimum width of the filename part and the
1621 * separators and this message, this message will "overflow"
1622 * making the line longer than the maximum width.
1623 */
1624 if (options->stat_width == -1)
1626 width = term_columns() - options->output_prefix_length;
1627 else
1628 width = options->stat_width ? options->stat_width : 80;
1629 number_width = decimal_width(max_change) > number_width ?
1630 decimal_width(max_change) : number_width;
1631 if (options->stat_graph_width == -1)
1633 options->stat_graph_width = diff_stat_graph_width;
1634 /*
1636 * Guarantee 3/8*16==6 for the graph part
1637 * and 5/8*16==10 for the filename part
1638 */
1639 if (width < 16 + 6 + number_width)
1640 width = 16 + 6 + number_width;
1641 /*
1643 * First assign sizes that are wanted, ignoring available width.
1644 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1645 * starting from "XXX" should fit in graph_width.
1646 */
1647 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1648 if (options->stat_graph_width &&
1649 options->stat_graph_width < graph_width)
1650 graph_width = options->stat_graph_width;
1651 name_width = (options->stat_name_width > 0 &&
1653 options->stat_name_width < max_len) ?
1654 options->stat_name_width : max_len;
1655 /*
1657 * Adjust adjustable widths not to exceed maximum width
1658 */
1659 if (name_width + number_width + 6 + graph_width > width) {
1660 if (graph_width > width * 3/8 - number_width - 6) {
1661 graph_width = width * 3/8 - number_width - 6;
1662 if (graph_width < 6)
1663 graph_width = 6;
1664 }
1665 if (options->stat_graph_width &&
1667 graph_width > options->stat_graph_width)
1668 graph_width = options->stat_graph_width;
1669 if (name_width > width - number_width - 6 - graph_width)
1670 name_width = width - number_width - 6 - graph_width;
1671 else
1672 graph_width = width - number_width - 6 - name_width;
1673 }
1674 /*
1676 * From here name_width is the width of the name area,
1677 * and graph_width is the width of the graph area.
1678 * max_change is used to scale graph properly.
1679 */
1680 for (i = 0; i < count; i++) {
1681 const char *prefix = "";
1682 struct diffstat_file *file = data->files[i];
1683 char *name = file->print_name;
1684 uintmax_t added = file->added;
1685 uintmax_t deleted = file->deleted;
1686 int name_len;
1687 if (!file->is_interesting && (added + deleted == 0))
1689 continue;
1690 /*
1692 * "scale" the filename
1693 */
1694 len = name_width;
1695 name_len = strlen(name);
1696 if (name_width < name_len) {
1697 char *slash;
1698 prefix = "...";
1699 len -= 3;
1700 name += name_len - len;
1701 slash = strchr(name, '/');
1702 if (slash)
1703 name = slash;
1704 }
1705 if (file->is_binary) {
1707 fprintf(options->file, "%s", line_prefix);
1708 show_name(options->file, prefix, name, len);
1709 fprintf(options->file, " %*s", number_width, "Bin");
1710 if (!added && !deleted) {
1711 putc('\n', options->file);
1712 continue;
1713 }
1714 fprintf(options->file, " %s%"PRIuMAX"%s",
1715 del_c, deleted, reset);
1716 fprintf(options->file, " -> ");
1717 fprintf(options->file, "%s%"PRIuMAX"%s",
1718 add_c, added, reset);
1719 fprintf(options->file, " bytes");
1720 fprintf(options->file, "\n");
1721 continue;
1722 }
1723 else if (file->is_unmerged) {
1724 fprintf(options->file, "%s", line_prefix);
1725 show_name(options->file, prefix, name, len);
1726 fprintf(options->file, " Unmerged\n");
1727 continue;
1728 }
1729 /*
1731 * scale the add/delete
1732 */
1733 add = added;
1734 del = deleted;
1735 if (graph_width <= max_change) {
1737 int total = scale_linear(add + del, graph_width, max_change);
1738 if (total < 2 && add && del)
1739 /* width >= 2 due to the sanity check */
1740 total = 2;
1741 if (add < del) {
1742 add = scale_linear(add, graph_width, max_change);
1743 del = total - add;
1744 } else {
1745 del = scale_linear(del, graph_width, max_change);
1746 add = total - del;
1747 }
1748 }
1749 fprintf(options->file, "%s", line_prefix);
1750 show_name(options->file, prefix, name, len);
1751 fprintf(options->file, " %*"PRIuMAX"%s",
1752 number_width, added + deleted,
1753 added + deleted ? " " : "");
1754 show_graph(options->file, '+', add, add_c, reset);
1755 show_graph(options->file, '-', del, del_c, reset);
1756 fprintf(options->file, "\n");
1757 }
1758 for (i = 0; i < data->nr; i++) {
1760 struct diffstat_file *file = data->files[i];
1761 uintmax_t added = file->added;
1762 uintmax_t deleted = file->deleted;
1763 if (file->is_unmerged ||
1765 (!file->is_interesting && (added + deleted == 0))) {
1766 total_files--;
1767 continue;
1768 }
1769 if (!file->is_binary) {
1771 adds += added;
1772 dels += deleted;
1773 }
1774 if (i < count)
1775 continue;
1776 if (!extra_shown)
1777 fprintf(options->file, "%s ...\n", line_prefix);
1778 extra_shown = 1;
1779 }
1780 fprintf(options->file, "%s", line_prefix);
1781 print_stat_summary(options->file, total_files, adds, dels);
1782}
1783static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1785{
1786 int i, adds = 0, dels = 0, total_files = data->nr;
1787 if (data->nr == 0)
1789 return;
1790 for (i = 0; i < data->nr; i++) {
1792 int added = data->files[i]->added;
1793 int deleted= data->files[i]->deleted;
1794 if (data->files[i]->is_unmerged ||
1796 (!data->files[i]->is_interesting && (added + deleted == 0))) {
1797 total_files--;
1798 } else if (!data->files[i]->is_binary) { /* don't count bytes */
1799 adds += added;
1800 dels += deleted;
1801 }
1802 }
1803 fprintf(options->file, "%s", diff_line_prefix(options));
1804 print_stat_summary(options->file, total_files, adds, dels);
1805}
1806static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1808{
1809 int i;
1810 if (data->nr == 0)
1812 return;
1813 for (i = 0; i < data->nr; i++) {
1815 struct diffstat_file *file = data->files[i];
1816 fprintf(options->file, "%s", diff_line_prefix(options));
1818 if (file->is_binary)
1820 fprintf(options->file, "-\t-\t");
1821 else
1822 fprintf(options->file,
1823 "%"PRIuMAX"\t%"PRIuMAX"\t",
1824 file->added, file->deleted);
1825 if (options->line_termination) {
1826 fill_print_name(file);
1827 if (!file->is_renamed)
1828 write_name_quoted(file->name, options->file,
1829 options->line_termination);
1830 else {
1831 fputs(file->print_name, options->file);
1832 putc(options->line_termination, options->file);
1833 }
1834 } else {
1835 if (file->is_renamed) {
1836 putc('\0', options->file);
1837 write_name_quoted(file->from_name, options->file, '\0');
1838 }
1839 write_name_quoted(file->name, options->file, '\0');
1840 }
1841 }
1842}
1843struct dirstat_file {
1845 const char *name;
1846 unsigned long changed;
1847};
1848struct dirstat_dir {
1850 struct dirstat_file *files;
1851 int alloc, nr, permille, cumulative;
1852};
1853static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1855 unsigned long changed, const char *base, int baselen)
1856{
1857 unsigned long this_dir = 0;
1858 unsigned int sources = 0;
1859 const char *line_prefix = diff_line_prefix(opt);
1860 while (dir->nr) {
1862 struct dirstat_file *f = dir->files;
1863 int namelen = strlen(f->name);
1864 unsigned long this;
1865 char *slash;
1866 if (namelen < baselen)
1868 break;
1869 if (memcmp(f->name, base, baselen))
1870 break;
1871 slash = strchr(f->name + baselen, '/');
1872 if (slash) {
1873 int newbaselen = slash + 1 - f->name;
1874 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1875 sources++;
1876 } else {
1877 this = f->changed;
1878 dir->files++;
1879 dir->nr--;
1880 sources += 2;
1881 }
1882 this_dir += this;
1883 }
1884 /*
1886 * We don't report dirstat's for
1887 * - the top level
1888 * - or cases where everything came from a single directory
1889 * under this directory (sources == 1).
1890 */
1891 if (baselen && sources != 1) {
1892 if (this_dir) {
1893 int permille = this_dir * 1000 / changed;
1894 if (permille >= dir->permille) {
1895 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1896 permille / 10, permille % 10, baselen, base);
1897 if (!dir->cumulative)
1898 return 0;
1899 }
1900 }
1901 }
1902 return this_dir;
1903}
1904static int dirstat_compare(const void *_a, const void *_b)
1906{
1907 const struct dirstat_file *a = _a;
1908 const struct dirstat_file *b = _b;
1909 return strcmp(a->name, b->name);
1910}
1911static void show_dirstat(struct diff_options *options)
1913{
1914 int i;
1915 unsigned long changed;
1916 struct dirstat_dir dir;
1917 struct diff_queue_struct *q = &diff_queued_diff;
1918 dir.files = NULL;
1920 dir.alloc = 0;
1921 dir.nr = 0;
1922 dir.permille = options->dirstat_permille;
1923 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1924 changed = 0;
1926 for (i = 0; i < q->nr; i++) {
1927 struct diff_filepair *p = q->queue[i];
1928 const char *name;
1929 unsigned long copied, added, damage;
1930 int content_changed;
1931 name = p->two->path ? p->two->path : p->one->path;
1933 if (p->one->oid_valid && p->two->oid_valid)
1935 content_changed = oidcmp(&p->one->oid, &p->two->oid);
1936 else
1937 content_changed = 1;
1938 if (!content_changed) {
1940 /*
1941 * The SHA1 has not changed, so pre-/post-content is
1942 * identical. We can therefore skip looking at the
1943 * file contents altogether.
1944 */
1945 damage = 0;
1946 goto found_damage;
1947 }
1948 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1950 /*
1951 * In --dirstat-by-file mode, we don't really need to
1952 * look at the actual file contents at all.
1953 * The fact that the SHA1 changed is enough for us to
1954 * add this file to the list of results
1955 * (with each file contributing equal damage).
1956 */
1957 damage = 1;
1958 goto found_damage;
1959 }
1960 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1962 diff_populate_filespec(p->one, 0);
1963 diff_populate_filespec(p->two, 0);
1964 diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1965 &copied, &added);
1966 diff_free_filespec_data(p->one);
1967 diff_free_filespec_data(p->two);
1968 } else if (DIFF_FILE_VALID(p->one)) {
1969 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
1970 copied = added = 0;
1971 diff_free_filespec_data(p->one);
1972 } else if (DIFF_FILE_VALID(p->two)) {
1973 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
1974 copied = 0;
1975 added = p->two->size;
1976 diff_free_filespec_data(p->two);
1977 } else
1978 continue;
1979 /*
1981 * Original minus copied is the removed material,
1982 * added is the new material. They are both damages
1983 * made to the preimage.
1984 * If the resulting damage is zero, we know that
1985 * diffcore_count_changes() considers the two entries to
1986 * be identical, but since content_changed is true, we
1987 * know that there must have been _some_ kind of change,
1988 * so we force all entries to have damage > 0.
1989 */
1990 damage = (p->one->size - copied) + added;
1991 if (!damage)
1992 damage = 1;
1993found_damage:
1995 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1996 dir.files[dir.nr].name = name;
1997 dir.files[dir.nr].changed = damage;
1998 changed += damage;
1999 dir.nr++;
2000 }
2001 /* This can happen even with many files, if everything was renames */
2003 if (!changed)
2004 return;
2005 /* Show all directories with more than x% of the changes */
2007 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
2008 gather_dirstat(options, &dir, changed, "", 0);
2009}
2010static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
2012{
2013 int i;
2014 unsigned long changed;
2015 struct dirstat_dir dir;
2016 if (data->nr == 0)
2018 return;
2019 dir.files = NULL;
2021 dir.alloc = 0;
2022 dir.nr = 0;
2023 dir.permille = options->dirstat_permille;
2024 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2025 changed = 0;
2027 for (i = 0; i < data->nr; i++) {
2028 struct diffstat_file *file = data->files[i];
2029 unsigned long damage = file->added + file->deleted;
2030 if (file->is_binary)
2031 /*
2032 * binary files counts bytes, not lines. Must find some
2033 * way to normalize binary bytes vs. textual lines.
2034 * The following heuristic assumes that there are 64
2035 * bytes per "line".
2036 * This is stupid and ugly, but very cheap...
2037 */
2038 damage = (damage + 63) / 64;
2039 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2040 dir.files[dir.nr].name = file->name;
2041 dir.files[dir.nr].changed = damage;
2042 changed += damage;
2043 dir.nr++;
2044 }
2045 /* This can happen even with many files, if everything was renames */
2047 if (!changed)
2048 return;
2049 /* Show all directories with more than x% of the changes */
2051 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
2052 gather_dirstat(options, &dir, changed, "", 0);
2053}
2054static void free_diffstat_info(struct diffstat_t *diffstat)
2056{
2057 int i;
2058 for (i = 0; i < diffstat->nr; i++) {
2059 struct diffstat_file *f = diffstat->files[i];
2060 if (f->name != f->print_name)
2061 free(f->print_name);
2062 free(f->name);
2063 free(f->from_name);
2064 free(f);
2065 }
2066 free(diffstat->files);
2067}
2068struct checkdiff_t {
2070 const char *filename;
2071 int lineno;
2072 int conflict_marker_size;
2073 struct diff_options *o;
2074 unsigned ws_rule;
2075 unsigned status;
2076};
2077static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2079{
2080 char firstchar;
2081 int cnt;
2082 if (len < marker_size + 1)
2084 return 0;
2085 firstchar = line[0];
2086 switch (firstchar) {
2087 case '=': case '>': case '<': case '|':
2088 break;
2089 default:
2090 return 0;
2091 }
2092 for (cnt = 1; cnt < marker_size; cnt++)
2093 if (line[cnt] != firstchar)
2094 return 0;
2095 /* line[1] thru line[marker_size-1] are same as firstchar */
2096 if (len < marker_size + 1 || !isspace(line[marker_size]))
2097 return 0;
2098 return 1;
2099}
2100static void checkdiff_consume(void *priv, char *line, unsigned long len)
2102{
2103 struct checkdiff_t *data = priv;
2104 int marker_size = data->conflict_marker_size;
2105 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2106 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2107 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2108 char *err;
2109 const char *line_prefix;
2110 assert(data->o);
2112 line_prefix = diff_line_prefix(data->o);
2113 if (line[0] == '+') {
2115 unsigned bad;
2116 data->lineno++;
2117 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2118 data->status |= 1;
2119 fprintf(data->o->file,
2120 "%s%s:%d: leftover conflict marker\n",
2121 line_prefix, data->filename, data->lineno);
2122 }
2123 bad = ws_check(line + 1, len - 1, data->ws_rule);
2124 if (!bad)
2125 return;
2126 data->status |= bad;
2127 err = whitespace_error_string(bad);
2128 fprintf(data->o->file, "%s%s:%d: %s.\n",
2129 line_prefix, data->filename, data->lineno, err);
2130 free(err);
2131 emit_line(data->o, set, reset, line, 1);
2132 ws_check_emit(line + 1, len - 1, data->ws_rule,
2133 data->o->file, set, reset, ws);
2134 } else if (line[0] == ' ') {
2135 data->lineno++;
2136 } else if (line[0] == '@') {
2137 char *plus = strchr(line, '+');
2138 if (plus)
2139 data->lineno = strtol(plus, NULL, 10) - 1;
2140 else
2141 die("invalid diff");
2142 }
2143}
2144static unsigned char *deflate_it(char *data,
2146 unsigned long size,
2147 unsigned long *result_size)
2148{
2149 int bound;
2150 unsigned char *deflated;
2151 git_zstream stream;
2152 git_deflate_init(&stream, zlib_compression_level);
2154 bound = git_deflate_bound(&stream, size);
2155 deflated = xmalloc(bound);
2156 stream.next_out = deflated;
2157 stream.avail_out = bound;
2158 stream.next_in = (unsigned char *)data;
2160 stream.avail_in = size;
2161 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2162 ; /* nothing */
2163 git_deflate_end(&stream);
2164 *result_size = stream.total_out;
2165 return deflated;
2166}
2167static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2169 const char *prefix)
2170{
2171 void *cp;
2172 void *delta;
2173 void *deflated;
2174 void *data;
2175 unsigned long orig_size;
2176 unsigned long delta_size;
2177 unsigned long deflate_size;
2178 unsigned long data_size;
2179 /* We could do deflated delta, or we could do just deflated two,
2181 * whichever is smaller.
2182 */
2183 delta = NULL;
2184 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2185 if (one->size && two->size) {
2186 delta = diff_delta(one->ptr, one->size,
2187 two->ptr, two->size,
2188 &delta_size, deflate_size);
2189 if (delta) {
2190 void *to_free = delta;
2191 orig_size = delta_size;
2192 delta = deflate_it(delta, delta_size, &delta_size);
2193 free(to_free);
2194 }
2195 }
2196 if (delta && delta_size < deflate_size) {
2198 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2199 free(deflated);
2200 data = delta;
2201 data_size = delta_size;
2202 }
2203 else {
2204 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2205 free(delta);
2206 data = deflated;
2207 data_size = deflate_size;
2208 }
2209 /* emit data encoded in base85 */
2211 cp = data;
2212 while (data_size) {
2213 int bytes = (52 < data_size) ? 52 : data_size;
2214 char line[70];
2215 data_size -= bytes;
2216 if (bytes <= 26)
2217 line[0] = bytes + 'A' - 1;
2218 else
2219 line[0] = bytes - 26 + 'a' - 1;
2220 encode_85(line + 1, cp, bytes);
2221 cp = (char *) cp + bytes;
2222 fprintf(file, "%s", prefix);
2223 fputs(line, file);
2224 fputc('\n', file);
2225 }
2226 fprintf(file, "%s\n", prefix);
2227 free(data);
2228}
2229static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2231 const char *prefix)
2232{
2233 fprintf(file, "%sGIT binary patch\n", prefix);
2234 emit_binary_diff_body(file, one, two, prefix);
2235 emit_binary_diff_body(file, two, one, prefix);
2236}
2237int diff_filespec_is_binary(struct diff_filespec *one)
2239{
2240 if (one->is_binary == -1) {
2241 diff_filespec_load_driver(one);
2242 if (one->driver->binary != -1)
2243 one->is_binary = one->driver->binary;
2244 else {
2245 if (!one->data && DIFF_FILE_VALID(one))
2246 diff_populate_filespec(one, CHECK_BINARY);
2247 if (one->is_binary == -1 && one->data)
2248 one->is_binary = buffer_is_binary(one->data,
2249 one->size);
2250 if (one->is_binary == -1)
2251 one->is_binary = 0;
2252 }
2253 }
2254 return one->is_binary;
2255}
2256static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2258{
2259 diff_filespec_load_driver(one);
2260 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2261}
2262void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2264{
2265 if (!options->a_prefix)
2266 options->a_prefix = a;
2267 if (!options->b_prefix)
2268 options->b_prefix = b;
2269}
2270struct userdiff_driver *get_textconv(struct diff_filespec *one)
2272{
2273 if (!DIFF_FILE_VALID(one))
2274 return NULL;
2275 diff_filespec_load_driver(one);
2277 return userdiff_get_textconv(one->driver);
2278}
2279static void builtin_diff(const char *name_a,
2281 const char *name_b,
2282 struct diff_filespec *one,
2283 struct diff_filespec *two,
2284 const char *xfrm_msg,
2285 int must_show_header,
2286 struct diff_options *o,
2287 int complete_rewrite)
2288{
2289 mmfile_t mf1, mf2;
2290 const char *lbl[2];
2291 char *a_one, *b_two;
2292 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2293 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2294 const char *a_prefix, *b_prefix;
2295 struct userdiff_driver *textconv_one = NULL;
2296 struct userdiff_driver *textconv_two = NULL;
2297 struct strbuf header = STRBUF_INIT;
2298 const char *line_prefix = diff_line_prefix(o);
2299 if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2301 (!one->mode || S_ISGITLINK(one->mode)) &&
2302 (!two->mode || S_ISGITLINK(two->mode))) {
2303 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2304 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2305 show_submodule_summary(o->file, one->path ? one->path : two->path,
2306 line_prefix,
2307 one->oid.hash, two->oid.hash,
2308 two->dirty_submodule,
2309 meta, del, add, reset);
2310 return;
2311 }
2312 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2314 textconv_one = get_textconv(one);
2315 textconv_two = get_textconv(two);
2316 }
2317 diff_set_mnemonic_prefix(o, "a/", "b/");
2319 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2320 a_prefix = o->b_prefix;
2321 b_prefix = o->a_prefix;
2322 } else {
2323 a_prefix = o->a_prefix;
2324 b_prefix = o->b_prefix;
2325 }
2326 /* Never use a non-valid filename anywhere if at all possible */
2328 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2329 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2330 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2332 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2333 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2334 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2335 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2336 if (lbl[0][0] == '/') {
2337 /* /dev/null */
2338 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2339 if (xfrm_msg)
2340 strbuf_addstr(&header, xfrm_msg);
2341 must_show_header = 1;
2342 }
2343 else if (lbl[1][0] == '/') {
2344 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2345 if (xfrm_msg)
2346 strbuf_addstr(&header, xfrm_msg);
2347 must_show_header = 1;
2348 }
2349 else {
2350 if (one->mode != two->mode) {
2351 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2352 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2353 must_show_header = 1;
2354 }
2355 if (xfrm_msg)
2356 strbuf_addstr(&header, xfrm_msg);
2357 /*
2359 * we do not run diff between different kind
2360 * of objects.
2361 */
2362 if ((one->mode ^ two->mode) & S_IFMT)
2363 goto free_ab_and_return;
2364 if (complete_rewrite &&
2365 (textconv_one || !diff_filespec_is_binary(one)) &&
2366 (textconv_two || !diff_filespec_is_binary(two))) {
2367 fprintf(o->file, "%s", header.buf);
2368 strbuf_reset(&header);
2369 emit_rewrite_diff(name_a, name_b, one, two,
2370 textconv_one, textconv_two, o);
2371 o->found_changes = 1;
2372 goto free_ab_and_return;
2373 }
2374 }
2375 if (o->irreversible_delete && lbl[1][0] == '/') {
2377 fprintf(o->file, "%s", header.buf);
2378 strbuf_reset(&header);
2379 goto free_ab_and_return;
2380 } else if (!DIFF_OPT_TST(o, TEXT) &&
2381 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2382 (!textconv_two && diff_filespec_is_binary(two)) )) {
2383 if (!one->data && !two->data &&
2384 S_ISREG(one->mode) && S_ISREG(two->mode) &&
2385 !DIFF_OPT_TST(o, BINARY)) {
2386 if (!oidcmp(&one->oid, &two->oid)) {
2387 if (must_show_header)
2388 fprintf(o->file, "%s", header.buf);
2389 goto free_ab_and_return;
2390 }
2391 fprintf(o->file, "%s", header.buf);
2392 fprintf(o->file, "%sBinary files %s and %s differ\n",
2393 line_prefix, lbl[0], lbl[1]);
2394 goto free_ab_and_return;
2395 }
2396 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2397 die("unable to read files to diff");
2398 /* Quite common confusing case */
2399 if (mf1.size == mf2.size &&
2400 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2401 if (must_show_header)
2402 fprintf(o->file, "%s", header.buf);
2403 goto free_ab_and_return;
2404 }
2405 fprintf(o->file, "%s", header.buf);
2406 strbuf_reset(&header);
2407 if (DIFF_OPT_TST(o, BINARY))
2408 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2409 else
2410 fprintf(o->file, "%sBinary files %s and %s differ\n",
2411 line_prefix, lbl[0], lbl[1]);
2412 o->found_changes = 1;
2413 } else {
2414 /* Crazy xdl interfaces.. */
2415 const char *diffopts = getenv("GIT_DIFF_OPTS");
2416 const char *v;
2417 xpparam_t xpp;
2418 xdemitconf_t xecfg;
2419 struct emit_callback ecbdata;
2420 const struct userdiff_funcname *pe;
2421 if (must_show_header) {
2423 fprintf(o->file, "%s", header.buf);
2424 strbuf_reset(&header);
2425 }
2426 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2428 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2429 pe = diff_funcname_pattern(one);
2431 if (!pe)
2432 pe = diff_funcname_pattern(two);
2433 memset(&xpp, 0, sizeof(xpp));
2435 memset(&xecfg, 0, sizeof(xecfg));
2436 memset(&ecbdata, 0, sizeof(ecbdata));
2437 ecbdata.label_path = lbl;
2438 ecbdata.color_diff = want_color(o->use_color);
2439 ecbdata.ws_rule = whitespace_rule(name_b);
2440 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2441 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2442 ecbdata.opt = o;
2443 ecbdata.header = header.len ? &header : NULL;
2444 xpp.flags = o->xdl_opts;
2445 xecfg.ctxlen = o->context;
2446 xecfg.interhunkctxlen = o->interhunkcontext;
2447 xecfg.flags = XDL_EMIT_FUNCNAMES;
2448 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2449 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2450 if (pe)
2451 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2452 if (!diffopts)
2453 ;
2454 else if (skip_prefix(diffopts, "--unified=", &v))
2455 xecfg.ctxlen = strtoul(v, NULL, 10);
2456 else if (skip_prefix(diffopts, "-u", &v))
2457 xecfg.ctxlen = strtoul(v, NULL, 10);
2458 if (o->word_diff)
2459 init_diff_words_data(&ecbdata, o, one, two);
2460 if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2461 &xpp, &xecfg))
2462 die("unable to generate diff for %s", one->path);
2463 if (o->word_diff)
2464 free_diff_words_data(&ecbdata);
2465 if (textconv_one)
2466 free(mf1.ptr);
2467 if (textconv_two)
2468 free(mf2.ptr);
2469 xdiff_clear_find_func(&xecfg);
2470 }
2471 free_ab_and_return:
2473 strbuf_release(&header);
2474 diff_free_filespec_data(one);
2475 diff_free_filespec_data(two);
2476 free(a_one);
2477 free(b_two);
2478 return;
2479}
2480static void builtin_diffstat(const char *name_a, const char *name_b,
2482 struct diff_filespec *one,
2483 struct diff_filespec *two,
2484 struct diffstat_t *diffstat,
2485 struct diff_options *o,
2486 struct diff_filepair *p)
2487{
2488 mmfile_t mf1, mf2;
2489 struct diffstat_file *data;
2490 int same_contents;
2491 int complete_rewrite = 0;
2492 if (!DIFF_PAIR_UNMERGED(p)) {
2494 if (p->status == DIFF_STATUS_MODIFIED && p->score)
2495 complete_rewrite = 1;
2496 }
2497 data = diffstat_add(diffstat, name_a, name_b);
2499 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2500 if (!one || !two) {
2502 data->is_unmerged = 1;
2503 return;
2504 }
2505 same_contents = !oidcmp(&one->oid, &two->oid);
2507 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2509 data->is_binary = 1;
2510 if (same_contents) {
2511 data->added = 0;
2512 data->deleted = 0;
2513 } else {
2514 data->added = diff_filespec_size(two);
2515 data->deleted = diff_filespec_size(one);
2516 }
2517 }
2518 else if (complete_rewrite) {
2520 diff_populate_filespec(one, 0);
2521 diff_populate_filespec(two, 0);
2522 data->deleted = count_lines(one->data, one->size);
2523 data->added = count_lines(two->data, two->size);
2524 }
2525 else if (!same_contents) {
2527 /* Crazy xdl interfaces.. */
2528 xpparam_t xpp;
2529 xdemitconf_t xecfg;
2530 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2532 die("unable to read files to diff");
2533 memset(&xpp, 0, sizeof(xpp));
2535 memset(&xecfg, 0, sizeof(xecfg));
2536 xpp.flags = o->xdl_opts;
2537 xecfg.ctxlen = o->context;
2538 xecfg.interhunkctxlen = o->interhunkcontext;
2539 if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2540 &xpp, &xecfg))
2541 die("unable to generate diffstat for %s", one->path);
2542 }
2543 diff_free_filespec_data(one);
2545 diff_free_filespec_data(two);
2546}
2547static void builtin_checkdiff(const char *name_a, const char *name_b,
2549 const char *attr_path,
2550 struct diff_filespec *one,
2551 struct diff_filespec *two,
2552 struct diff_options *o)
2553{
2554 mmfile_t mf1, mf2;
2555 struct checkdiff_t data;
2556 if (!two)
2558 return;
2559 memset(&data, 0, sizeof(data));
2561 data.filename = name_b ? name_b : name_a;
2562 data.lineno = 0;
2563 data.o = o;
2564 data.ws_rule = whitespace_rule(attr_path);
2565 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2566 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2568 die("unable to read files to diff");
2569 /*
2571 * All the other codepaths check both sides, but not checking
2572 * the "old" side here is deliberate. We are checking the newly
2573 * introduced changes, and as long as the "new" side is text, we
2574 * can and should check what it introduces.
2575 */
2576 if (diff_filespec_is_binary(two))
2577 goto free_and_return;
2578 else {
2579 /* Crazy xdl interfaces.. */
2580 xpparam_t xpp;
2581 xdemitconf_t xecfg;
2582 memset(&xpp, 0, sizeof(xpp));
2584 memset(&xecfg, 0, sizeof(xecfg));
2585 xecfg.ctxlen = 1; /* at least one context line */
2586 xpp.flags = 0;
2587 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2588 &xpp, &xecfg))
2589 die("unable to generate checkdiff for %s", one->path);
2590 if (data.ws_rule & WS_BLANK_AT_EOF) {
2592 struct emit_callback ecbdata;
2593 int blank_at_eof;
2594 ecbdata.ws_rule = data.ws_rule;
2596 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2597 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2598 if (blank_at_eof) {
2600 static char *err;
2601 if (!err)
2602 err = whitespace_error_string(WS_BLANK_AT_EOF);
2603 fprintf(o->file, "%s:%d: %s.\n",
2604 data.filename, blank_at_eof, err);
2605 data.status = 1; /* report errors */
2606 }
2607 }
2608 }
2609 free_and_return:
2610 diff_free_filespec_data(one);
2611 diff_free_filespec_data(two);
2612 if (data.status)
2613 DIFF_OPT_SET(o, CHECK_FAILED);
2614}
2615struct diff_filespec *alloc_filespec(const char *path)
2617{
2618 struct diff_filespec *spec;
2619 FLEXPTR_ALLOC_STR(spec, path, path);
2621 spec->count = 1;
2622 spec->is_binary = -1;
2623 return spec;
2624}
2625void free_filespec(struct diff_filespec *spec)
2627{
2628 if (!--spec->count) {
2629 diff_free_filespec_data(spec);
2630 free(spec);
2631 }
2632}
2633void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2635 int sha1_valid, unsigned short mode)
2636{
2637 if (mode) {
2638 spec->mode = canon_mode(mode);
2639 hashcpy(spec->oid.hash, sha1);
2640 spec->oid_valid = sha1_valid;
2641 }
2642}
2643/*
2645 * Given a name and sha1 pair, if the index tells us the file in
2646 * the work tree has that object contents, return true, so that
2647 * prepare_temp_file() does not have to inflate and extract.
2648 */
2649static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2650{
2651 const struct cache_entry *ce;
2652 struct stat st;
2653 int pos, len;
2654 /*
2656 * We do not read the cache ourselves here, because the
2657 * benchmark with my previous version that always reads cache
2658 * shows that it makes things worse for diff-tree comparing
2659 * two linux-2.6 kernel trees in an already checked out work
2660 * tree. This is because most diff-tree comparisons deal with
2661 * only a small number of files, while reading the cache is
2662 * expensive for a large project, and its cost outweighs the
2663 * savings we get by not inflating the object to a temporary
2664 * file. Practically, this code only helps when we are used
2665 * by diff-cache --cached, which does read the cache before
2666 * calling us.
2667 */
2668 if (!active_cache)
2669 return 0;
2670 /* We want to avoid the working directory if our caller
2672 * doesn't need the data in a normal file, this system
2673 * is rather slow with its stat/open/mmap/close syscalls,
2674 * and the object is contained in a pack file. The pack
2675 * is probably already open and will be faster to obtain
2676 * the data through than the working directory. Loose
2677 * objects however would tend to be slower as they need
2678 * to be individually opened and inflated.
2679 */
2680 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2681 return 0;
2682 /*
2684 * Similarly, if we'd have to convert the file contents anyway, that
2685 * makes the optimization not worthwhile.
2686 */
2687 if (!want_file && would_convert_to_git(name))
2688 return 0;
2689 len = strlen(name);
2691 pos = cache_name_pos(name, len);
2692 if (pos < 0)
2693 return 0;
2694 ce = active_cache[pos];
2695 /*
2697 * This is not the sha1 we are looking for, or
2698 * unreusable because it is not a regular file.
2699 */
2700 if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2701 return 0;
2702 /*
2704 * If ce is marked as "assume unchanged", there is no
2705 * guarantee that work tree matches what we are looking for.
2706 */
2707 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2708 return 0;
2709 /*
2711 * If ce matches the file in the work tree, we can reuse it.
2712 */
2713 if (ce_uptodate(ce) ||
2714 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2715 return 1;
2716 return 0;
2718}
2719static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2721{
2722 struct strbuf buf = STRBUF_INIT;
2723 char *dirty = "";
2724 /* Are we looking at the work tree? */
2726 if (s->dirty_submodule)
2727 dirty = "-dirty";
2728 strbuf_addf(&buf, "Subproject commit %s%s\n",
2730 oid_to_hex(&s->oid), dirty);
2731 s->size = buf.len;
2732 if (size_only) {
2733 s->data = NULL;
2734 strbuf_release(&buf);
2735 } else {
2736 s->data = strbuf_detach(&buf, NULL);
2737 s->should_free = 1;
2738 }
2739 return 0;
2740}
2741/*
2743 * While doing rename detection and pickaxe operation, we may need to
2744 * grab the data for the blob (or file) for our own in-core comparison.
2745 * diff_filespec has data and size fields for this purpose.
2746 */
2747int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
2748{
2749 int size_only = flags & CHECK_SIZE_ONLY;
2750 int err = 0;
2751 /*
2752 * demote FAIL to WARN to allow inspecting the situation
2753 * instead of refusing.
2754 */
2755 enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
2756 ? SAFE_CRLF_WARN
2757 : safe_crlf);
2758 if (!DIFF_FILE_VALID(s))
2760 die("internal error: asking to populate invalid file.");
2761 if (S_ISDIR(s->mode))
2762 return -1;
2763 if (s->data)
2765 return 0;
2766 if (size_only && 0 < s->size)
2768 return 0;
2769 if (S_ISGITLINK(s->mode))
2771 return diff_populate_gitlink(s, size_only);
2772 if (!s->oid_valid ||
2774 reuse_worktree_file(s->path, s->oid.hash, 0)) {
2775 struct strbuf buf = STRBUF_INIT;
2776 struct stat st;
2777 int fd;
2778 if (lstat(s->path, &st) < 0) {
2780 if (errno == ENOENT) {
2781 err_empty:
2782 err = -1;
2783 empty:
2784 s->data = (char *)"";
2785 s->size = 0;
2786 return err;
2787 }
2788 }
2789 s->size = xsize_t(st.st_size);
2790 if (!s->size)
2791 goto empty;
2792 if (S_ISLNK(st.st_mode)) {
2793 struct strbuf sb = STRBUF_INIT;
2794 if (strbuf_readlink(&sb, s->path, s->size))
2796 goto err_empty;
2797 s->size = sb.len;
2798 s->data = strbuf_detach(&sb, NULL);
2799 s->should_free = 1;
2800 return 0;
2801 }
2802 if (size_only)
2803 return 0;
2804 if ((flags & CHECK_BINARY) &&
2805 s->size > big_file_threshold && s->is_binary == -1) {
2806 s->is_binary = 1;
2807 return 0;
2808 }
2809 fd = open(s->path, O_RDONLY);
2810 if (fd < 0)
2811 goto err_empty;
2812 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2813 close(fd);
2814 s->should_munmap = 1;
2815 /*
2817 * Convert from working tree format to canonical git format
2818 */
2819 if (convert_to_git(s->path, s->data, s->size, &buf, crlf_warn)) {
2820 size_t size = 0;
2821 munmap(s->data, s->size);
2822 s->should_munmap = 0;
2823 s->data = strbuf_detach(&buf, &size);
2824 s->size = size;
2825 s->should_free = 1;
2826 }
2827 }
2828 else {
2829 enum object_type type;
2830 if (size_only || (flags & CHECK_BINARY)) {
2831 type = sha1_object_info(s->oid.hash, &s->size);
2832 if (type < 0)
2833 die("unable to read %s",
2834 oid_to_hex(&s->oid));
2835 if (size_only)
2836 return 0;
2837 if (s->size > big_file_threshold && s->is_binary == -1) {
2838 s->is_binary = 1;
2839 return 0;
2840 }
2841 }
2842 s->data = read_sha1_file(s->oid.hash, &type, &s->size);
2843 if (!s->data)
2844 die("unable to read %s", oid_to_hex(&s->oid));
2845 s->should_free = 1;
2846 }
2847 return 0;
2848}
2849void diff_free_filespec_blob(struct diff_filespec *s)
2851{
2852 if (s->should_free)
2853 free(s->data);
2854 else if (s->should_munmap)
2855 munmap(s->data, s->size);
2856 if (s->should_free || s->should_munmap) {
2858 s->should_free = s->should_munmap = 0;
2859 s->data = NULL;
2860 }
2861}
2862void diff_free_filespec_data(struct diff_filespec *s)
2864{
2865 diff_free_filespec_blob(s);
2866 free(s->cnt_data);
2867 s->cnt_data = NULL;
2868}
2869static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2871 void *blob,
2872 unsigned long size,
2873 const struct object_id *oid,
2874 int mode)
2875{
2876 int fd;
2877 struct strbuf buf = STRBUF_INIT;
2878 struct strbuf template = STRBUF_INIT;
2879 char *path_dup = xstrdup(path);
2880 const char *base = basename(path_dup);
2881 /* Generate "XXXXXX_basename.ext" */
2883 strbuf_addstr(&template, "XXXXXX_");
2884 strbuf_addstr(&template, base);
2885 fd = mks_tempfile_ts(&temp->tempfile, template.buf, strlen(base) + 1);
2887 if (fd < 0)
2888 die_errno("unable to create temp-file");
2889 if (convert_to_working_tree(path,
2890 (const char *)blob, (size_t)size, &buf)) {
2891 blob = buf.buf;
2892 size = buf.len;
2893 }
2894 if (write_in_full(fd, blob, size) != size)
2895 die_errno("unable to write temp-file");
2896 close_tempfile(&temp->tempfile);
2897 temp->name = get_tempfile_path(&temp->tempfile);
2898 oid_to_hex_r(temp->hex, oid);
2899 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", mode);
2900 strbuf_release(&buf);
2901 strbuf_release(&template);
2902 free(path_dup);
2903}
2904static struct diff_tempfile *prepare_temp_file(const char *name,
2906 struct diff_filespec *one)
2907{
2908 struct diff_tempfile *temp = claim_diff_tempfile();
2909 if (!DIFF_FILE_VALID(one)) {
2911 not_a_valid_file:
2912 /* A '-' entry produces this for file-2, and
2913 * a '+' entry produces this for file-1.
2914 */
2915 temp->name = "/dev/null";
2916 xsnprintf(temp->hex, sizeof(temp->hex), ".");
2917 xsnprintf(temp->mode, sizeof(temp->mode), ".");
2918 return temp;
2919 }
2920 if (!S_ISGITLINK(one->mode) &&
2922 (!one->oid_valid ||
2923 reuse_worktree_file(name, one->oid.hash, 1))) {
2924 struct stat st;
2925 if (lstat(name, &st) < 0) {
2926 if (errno == ENOENT)
2927 goto not_a_valid_file;
2928 die_errno("stat(%s)", name);
2929 }
2930 if (S_ISLNK(st.st_mode)) {
2931 struct strbuf sb = STRBUF_INIT;
2932 if (strbuf_readlink(&sb, name, st.st_size) < 0)
2933 die_errno("readlink(%s)", name);
2934 prep_temp_blob(name, temp, sb.buf, sb.len,
2935 (one->oid_valid ?
2936 &one->oid : &null_oid),
2937 (one->oid_valid ?
2938 one->mode : S_IFLNK));
2939 strbuf_release(&sb);
2940 }
2941 else {
2942 /* we can borrow from the file in the work tree */
2943 temp->name = name;
2944 if (!one->oid_valid)
2945 sha1_to_hex_r(temp->hex, null_sha1);
2946 else
2947 sha1_to_hex_r(temp->hex, one->oid.hash);
2948 /* Even though we may sometimes borrow the
2949 * contents from the work tree, we always want
2950 * one->mode. mode is trustworthy even when
2951 * !(one->sha1_valid), as long as
2952 * DIFF_FILE_VALID(one).
2953 */
2954 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", one->mode);
2955 }
2956 return temp;
2957 }
2958 else {
2959 if (diff_populate_filespec(one, 0))
2960 die("cannot read data blob for %s", one->path);
2961 prep_temp_blob(name, temp, one->data, one->size,
2962 &one->oid, one->mode);
2963 }
2964 return temp;
2965}
2966static void add_external_diff_name(struct argv_array *argv,
2968 const char *name,
2969 struct diff_filespec *df)
2970{
2971 struct diff_tempfile *temp = prepare_temp_file(name, df);
2972 argv_array_push(argv, temp->name);
2973 argv_array_push(argv, temp->hex);
2974 argv_array_push(argv, temp->mode);
2975}
2976/* An external diff command takes:
2978 *
2979 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2980 * infile2 infile2-sha1 infile2-mode [ rename-to ]
2981 *
2982 */
2983static void run_external_diff(const char *pgm,
2984 const char *name,
2985 const char *other,
2986 struct diff_filespec *one,
2987 struct diff_filespec *two,
2988 const char *xfrm_msg,
2989 int complete_rewrite,
2990 struct diff_options *o)
2991{
2992 struct argv_array argv = ARGV_ARRAY_INIT;
2993 struct argv_array env = ARGV_ARRAY_INIT;
2994 struct diff_queue_struct *q = &diff_queued_diff;
2995 argv_array_push(&argv, pgm);
2997 argv_array_push(&argv, name);
2998 if (one && two) {
3000 add_external_diff_name(&argv, name, one);
3001 if (!other)
3002 add_external_diff_name(&argv, name, two);
3003 else {
3004 add_external_diff_name(&argv, other, two);
3005 argv_array_push(&argv, other);
3006 argv_array_push(&argv, xfrm_msg);
3007 }
3008 }
3009 argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
3011 argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
3012 if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
3014 die(_("external diff died, stopping at %s"), name);
3015 remove_tempfile();
3017 argv_array_clear(&argv);
3018 argv_array_clear(&env);
3019}
3020static int similarity_index(struct diff_filepair *p)
3022{
3023 return p->score * 100 / MAX_SCORE;
3024}
3025static void fill_metainfo(struct strbuf *msg,
3027 const char *name,
3028 const char *other,
3029 struct diff_filespec *one,
3030 struct diff_filespec *two,
3031 struct diff_options *o,
3032 struct diff_filepair *p,
3033 int *must_show_header,
3034 int use_color)
3035{
3036 const char *set = diff_get_color(use_color, DIFF_METAINFO);
3037 const char *reset = diff_get_color(use_color, DIFF_RESET);
3038 const char *line_prefix = diff_line_prefix(o);
3039 *must_show_header = 1;
3041 strbuf_init(msg, PATH_MAX * 2 + 300);
3042 switch (p->status) {
3043 case DIFF_STATUS_COPIED:
3044 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3045 line_prefix, set, similarity_index(p));
3046 strbuf_addf(msg, "%s\n%s%scopy from ",
3047 reset, line_prefix, set);
3048 quote_c_style(name, msg, NULL, 0);
3049 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
3050 quote_c_style(other, msg, NULL, 0);
3051 strbuf_addf(msg, "%s\n", reset);
3052 break;
3053 case DIFF_STATUS_RENAMED:
3054 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3055 line_prefix, set, similarity_index(p));
3056 strbuf_addf(msg, "%s\n%s%srename from ",
3057 reset, line_prefix, set);
3058 quote_c_style(name, msg, NULL, 0);
3059 strbuf_addf(msg, "%s\n%s%srename to ",
3060 reset, line_prefix, set);
3061 quote_c_style(other, msg, NULL, 0);
3062 strbuf_addf(msg, "%s\n", reset);
3063 break;
3064 case DIFF_STATUS_MODIFIED:
3065 if (p->score) {
3066 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
3067 line_prefix,
3068 set, similarity_index(p), reset);
3069 break;
3070 }
3071 /* fallthru */
3072 default:
3073 *must_show_header = 0;
3074 }
3075 if (one && two && oidcmp(&one->oid, &two->oid)) {
3076 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
3077 if (DIFF_OPT_TST(o, BINARY)) {
3079 mmfile_t mf;
3080 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3081 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3082 abbrev = 40;
3083 }
3084 strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
3085 find_unique_abbrev(one->oid.hash, abbrev));
3086 strbuf_addstr(msg, find_unique_abbrev(two->oid.hash, abbrev));
3087 if (one->mode == two->mode)
3088 strbuf_addf(msg, " %06o", one->mode);
3089 strbuf_addf(msg, "%s\n", reset);
3090 }
3091}
3092static void run_diff_cmd(const char *pgm,
3094 const char *name,
3095 const char *other,
3096 const char *attr_path,
3097 struct diff_filespec *one,
3098 struct diff_filespec *two,
3099 struct strbuf *msg,
3100 struct diff_options *o,
3101 struct diff_filepair *p)
3102{
3103 const char *xfrm_msg = NULL;
3104 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3105 int must_show_header = 0;
3106 if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3109 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3110 if (drv && drv->external)
3111 pgm = drv->external;
3112 }
3113 if (msg) {
3115 /*
3116 * don't use colors when the header is intended for an
3117 * external diff driver
3118 */
3119 fill_metainfo(msg, name, other, one, two, o, p,
3120 &must_show_header,
3121 want_color(o->use_color) && !pgm);
3122 xfrm_msg = msg->len ? msg->buf : NULL;
3123 }
3124 if (pgm) {
3126 run_external_diff(pgm, name, other, one, two, xfrm_msg,
3127 complete_rewrite, o);
3128 return;
3129 }
3130 if (one && two)
3131 builtin_diff(name, other ? other : name,
3132 one, two, xfrm_msg, must_show_header,
3133 o, complete_rewrite);
3134 else
3135 fprintf(o->file, "* Unmerged path %s\n", name);
3136}
3137static void diff_fill_sha1_info(struct diff_filespec *one)
3139{
3140 if (DIFF_FILE_VALID(one)) {
3141 if (!one->oid_valid) {
3142 struct stat st;
3143 if (one->is_stdin) {
3144 oidclr(&one->oid);
3145 return;
3146 }
3147 if (lstat(one->path, &st) < 0)
3148 die_errno("stat '%s'", one->path);
3149 if (index_path(one->oid.hash, one->path, &st, 0))
3150 die("cannot hash %s", one->path);
3151 }
3152 }
3153 else
3154 oidclr(&one->oid);
3155}
3156static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3158{
3159 /* Strip the prefix but do not molest /dev/null and absolute paths */
3160 if (*namep && **namep != '/') {
3161 *namep += prefix_length;
3162 if (**namep == '/')
3163 ++*namep;
3164 }
3165 if (*otherp && **otherp != '/') {
3166 *otherp += prefix_length;
3167 if (**otherp == '/')
3168 ++*otherp;
3169 }
3170}
3171static void run_diff(struct diff_filepair *p, struct diff_options *o)
3173{
3174 const char *pgm = external_diff();
3175 struct strbuf msg;
3176 struct diff_filespec *one = p->one;
3177 struct diff_filespec *two = p->two;
3178 const char *name;
3179 const char *other;
3180 const char *attr_path;
3181 name = p->one->path;
3183 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3184 attr_path = name;
3185 if (o->prefix_length)
3186 strip_prefix(o->prefix_length, &name, &other);
3187 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3189 pgm = NULL;
3190 if (DIFF_PAIR_UNMERGED(p)) {
3192 run_diff_cmd(pgm, name, NULL, attr_path,
3193 NULL, NULL, NULL, o, p);
3194 return;
3195 }
3196 diff_fill_sha1_info(one);
3198 diff_fill_sha1_info(two);
3199 if (!pgm &&
3201 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3202 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3203 /*
3204 * a filepair that changes between file and symlink
3205 * needs to be split into deletion and creation.
3206 */
3207 struct diff_filespec *null = alloc_filespec(two->path);
3208 run_diff_cmd(NULL, name, other, attr_path,
3209 one, null, &msg, o, p);
3210 free(null);
3211 strbuf_release(&msg);
3212 null = alloc_filespec(one->path);
3214 run_diff_cmd(NULL, name, other, attr_path,
3215 null, two, &msg, o, p);
3216 free(null);
3217 }
3218 else
3219 run_diff_cmd(pgm, name, other, attr_path,
3220 one, two, &msg, o, p);
3221 strbuf_release(&msg);
3223}
3224static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3226 struct diffstat_t *diffstat)
3227{
3228 const char *name;
3229 const char *other;
3230 if (DIFF_PAIR_UNMERGED(p)) {
3232 /* unmerged */
3233 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3234 return;
3235 }
3236 name = p->one->path;
3238 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3239 if (o->prefix_length)
3241 strip_prefix(o->prefix_length, &name, &other);
3242 diff_fill_sha1_info(p->one);
3244 diff_fill_sha1_info(p->two);
3245 builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3247}
3248static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3250{
3251 const char *name;
3252 const char *other;
3253 const char *attr_path;
3254 if (DIFF_PAIR_UNMERGED(p)) {
3256 /* unmerged */
3257 return;
3258 }
3259 name = p->one->path;
3261 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3262 attr_path = other ? other : name;
3263 if (o->prefix_length)
3265 strip_prefix(o->prefix_length, &name, &other);
3266 diff_fill_sha1_info(p->one);
3268 diff_fill_sha1_info(p->two);
3269 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3271}
3272void diff_setup(struct diff_options *options)
3274{
3275 memcpy(options, &default_diff_options, sizeof(*options));
3276 options->file = stdout;
3278 options->line_termination = '\n';
3280 options->break_opt = -1;
3281 options->rename_limit = -1;
3282 options->dirstat_permille = diff_dirstat_permille_default;
3283 options->context = diff_context_default;
3284 options->ws_error_highlight = WSEH_NEW;
3285 DIFF_OPT_SET(options, RENAME_EMPTY);
3286 /* pathchange left =NULL by default */
3288 options->change = diff_change;
3289 options->add_remove = diff_addremove;
3290 options->use_color = diff_use_color_default;
3291 options->detect_rename = diff_detect_rename_default;
3292 options->xdl_opts |= diff_algorithm;
3293 if (diff_compaction_heuristic)
3294 DIFF_XDL_SET(options, COMPACTION_HEURISTIC);
3295 options->orderfile = diff_order_file_cfg;
3297 if (diff_no_prefix) {
3299 options->a_prefix = options->b_prefix = "";
3300 } else if (!diff_mnemonic_prefix) {
3301 options->a_prefix = "a/";
3302 options->b_prefix = "b/";
3303 }
3304}
3305void diff_setup_done(struct diff_options *options)
3307{
3308 int count = 0;
3309 if (options->set_default)
3311 options->set_default(options);
3312 if (options->output_format & DIFF_FORMAT_NAME)
3314 count++;
3315 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3316 count++;
3317 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3318 count++;
3319 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3320 count++;
3321 if (count > 1)
3322 die("--name-only, --name-status, --check and -s are mutually exclusive");
3323 /*
3325 * Most of the time we can say "there are changes"
3326 * only by checking if there are changed paths, but
3327 * --ignore-whitespace* options force us to look
3328 * inside contents.
3329 */
3330 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3332 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3333 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3334 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3335 else
3336 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3337 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3339 options->detect_rename = DIFF_DETECT_COPY;
3340 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3342 options->prefix = NULL;
3343 if (options->prefix)
3344 options->prefix_length = strlen(options->prefix);
3345 else
3346 options->prefix_length = 0;
3347 if (options->output_format & (DIFF_FORMAT_NAME |
3349 DIFF_FORMAT_NAME_STATUS |
3350 DIFF_FORMAT_CHECKDIFF |
3351 DIFF_FORMAT_NO_OUTPUT))
3352 options->output_format &= ~(DIFF_FORMAT_RAW |
3353 DIFF_FORMAT_NUMSTAT |
3354 DIFF_FORMAT_DIFFSTAT |
3355 DIFF_FORMAT_SHORTSTAT |
3356 DIFF_FORMAT_DIRSTAT |
3357 DIFF_FORMAT_SUMMARY |
3358 DIFF_FORMAT_PATCH);
3359 /*
3361 * These cases always need recursive; we do not drop caller-supplied
3362 * recursive bits for other formats here.
3363 */
3364 if (options->output_format & (DIFF_FORMAT_PATCH |
3365 DIFF_FORMAT_NUMSTAT |
3366 DIFF_FORMAT_DIFFSTAT |
3367 DIFF_FORMAT_SHORTSTAT |
3368 DIFF_FORMAT_DIRSTAT |
3369 DIFF_FORMAT_SUMMARY |
3370 DIFF_FORMAT_CHECKDIFF))
3371 DIFF_OPT_SET(options, RECURSIVE);
3372 /*
3373 * Also pickaxe would not work very well if you do not say recursive
3374 */
3375 if (options->pickaxe)
3376 DIFF_OPT_SET(options, RECURSIVE);
3377 /*
3378 * When patches are generated, submodules diffed against the work tree
3379 * must be checked for dirtiness too so it can be shown in the output
3380 */
3381 if (options->output_format & DIFF_FORMAT_PATCH)
3382 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3383 if (options->detect_rename && options->rename_limit < 0)
3385 options->rename_limit = diff_rename_limit_default;
3386 if (options->setup & DIFF_SETUP_USE_CACHE) {
3387 if (!active_cache)
3388 /* read-cache does not die even when it fails
3389 * so it is safe for us to do this here. Also
3390 * it does not smudge active_cache or active_nr
3391 * when it fails, so we do not have to worry about
3392 * cleaning it up ourselves either.
3393 */
3394 read_cache();
3395 }
3396 if (options->abbrev <= 0 || 40 < options->abbrev)
3397 options->abbrev = 40; /* full */
3398 /*
3400 * It does not make sense to show the first hit we happened
3401 * to have found. It does not make sense not to return with
3402 * exit code in such a case either.
3403 */
3404 if (DIFF_OPT_TST(options, QUICK)) {
3405 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3406 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3407 }
3408 options->diff_path_counter = 0;
3410 if (DIFF_OPT_TST(options, FOLLOW_RENAMES) && options->pathspec.nr != 1)
3412 die(_("--follow requires exactly one pathspec"));
3413}
3414static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3416{
3417 char c, *eq;
3418 int len;
3419 if (*arg != '-')
3421 return 0;
3422 c = *++arg;
3423 if (!c)
3424 return 0;
3425 if (c == arg_short) {
3426 c = *++arg;
3427 if (!c)
3428 return 1;
3429 if (val && isdigit(c)) {
3430 char *end;
3431 int n = strtoul(arg, &end, 10);
3432 if (*end)
3433 return 0;
3434 *val = n;
3435 return 1;
3436 }
3437 return 0;
3438 }
3439 if (c != '-')
3440 return 0;
3441 arg++;
3442 eq = strchrnul(arg, '=');
3443 len = eq - arg;
3444 if (!len || strncmp(arg, arg_long, len))
3445 return 0;
3446 if (*eq) {
3447 int n;
3448 char *end;
3449 if (!isdigit(*++eq))
3450 return 0;
3451 n = strtoul(eq, &end, 10);
3452 if (*end)
3453 return 0;
3454 *val = n;
3455 }
3456 return 1;
3457}
3458static int diff_scoreopt_parse(const char *opt);
3460static inline int short_opt(char opt, const char **argv,
3462 const char **optarg)
3463{
3464 const char *arg = argv[0];
3465 if (arg[0] != '-' || arg[1] != opt)
3466 return 0;
3467 if (arg[2] != '\0') {
3468 *optarg = arg + 2;
3469 return 1;
3470 }
3471 if (!argv[1])
3472 die("Option '%c' requires a value", opt);
3473 *optarg = argv[1];
3474 return 2;
3475}
3476int parse_long_opt(const char *opt, const char **argv,
3478 const char **optarg)
3479{
3480 const char *arg = argv[0];
3481 if (!skip_prefix(arg, "--", &arg))
3482 return 0;
3483 if (!skip_prefix(arg, opt, &arg))
3484 return 0;
3485 if (*arg == '=') { /* stuck form: --option=value */
3486 *optarg = arg + 1;
3487 return 1;
3488 }
3489 if (*arg != '\0')
3490 return 0;
3491 /* separate form: --option value */
3492 if (!argv[1])
3493 die("Option '--%s' requires a value", opt);
3494 *optarg = argv[1];
3495 return 2;
3496}
3497static int stat_opt(struct diff_options *options, const char **av)
3499{
3500 const char *arg = av[0];
3501 char *end;
3502 int width = options->stat_width;
3503 int name_width = options->stat_name_width;
3504 int graph_width = options->stat_graph_width;
3505 int count = options->stat_count;
3506 int argcount = 1;
3507 if (!skip_prefix(arg, "--stat", &arg))
3509 die("BUG: stat option does not begin with --stat: %s", arg);
3510 end = (char *)arg;
3511 switch (*arg) {
3513 case '-':
3514 if (skip_prefix(arg, "-width", &arg)) {
3515 if (*arg == '=')
3516 width = strtoul(arg + 1, &end, 10);
3517 else if (!*arg && !av[1])
3518 die("Option '--stat-width' requires a value");
3519 else if (!*arg) {
3520 width = strtoul(av[1], &end, 10);
3521 argcount = 2;
3522 }
3523 } else if (skip_prefix(arg, "-name-width", &arg)) {
3524 if (*arg == '=')
3525 name_width = strtoul(arg + 1, &end, 10);
3526 else if (!*arg && !av[1])
3527 die("Option '--stat-name-width' requires a value");
3528 else if (!*arg) {
3529 name_width = strtoul(av[1], &end, 10);
3530 argcount = 2;
3531 }
3532 } else if (skip_prefix(arg, "-graph-width", &arg)) {
3533 if (*arg == '=')
3534 graph_width = strtoul(arg + 1, &end, 10);
3535 else if (!*arg && !av[1])
3536 die("Option '--stat-graph-width' requires a value");
3537 else if (!*arg) {
3538 graph_width = strtoul(av[1], &end, 10);
3539 argcount = 2;
3540 }
3541 } else if (skip_prefix(arg, "-count", &arg)) {
3542 if (*arg == '=')
3543 count = strtoul(arg + 1, &end, 10);
3544 else if (!*arg && !av[1])
3545 die("Option '--stat-count' requires a value");
3546 else if (!*arg) {
3547 count = strtoul(av[1], &end, 10);
3548 argcount = 2;
3549 }
3550 }
3551 break;
3552 case '=':
3553 width = strtoul(arg+1, &end, 10);
3554 if (*end == ',')
3555 name_width = strtoul(end+1, &end, 10);
3556 if (*end == ',')
3557 count = strtoul(end+1, &end, 10);
3558 }
3559 /* Important! This checks all the error cases! */
3561 if (*end)
3562 return 0;
3563 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3564 options->stat_name_width = name_width;
3565 options->stat_graph_width = graph_width;
3566 options->stat_width = width;
3567 options->stat_count = count;
3568 return argcount;
3569}
3570static int parse_dirstat_opt(struct diff_options *options, const char *params)
3572{
3573 struct strbuf errmsg = STRBUF_INIT;
3574 if (parse_dirstat_params(options, params, &errmsg))
3575 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3576 errmsg.buf);
3577 strbuf_release(&errmsg);
3578 /*
3579 * The caller knows a dirstat-related option is given from the command
3580 * line; allow it to say "return this_function();"
3581 */
3582 options->output_format |= DIFF_FORMAT_DIRSTAT;
3583 return 1;
3584}
3585static int parse_submodule_opt(struct diff_options *options, const char *value)
3587{
3588 if (parse_submodule_params(options, value))
3589 die(_("Failed to parse --submodule option parameter: '%s'"),
3590 value);
3591 return 1;
3592}
3593static const char diff_status_letters[] = {
3595 DIFF_STATUS_ADDED,
3596 DIFF_STATUS_COPIED,
3597 DIFF_STATUS_DELETED,
3598 DIFF_STATUS_MODIFIED,
3599 DIFF_STATUS_RENAMED,
3600 DIFF_STATUS_TYPE_CHANGED,
3601 DIFF_STATUS_UNKNOWN,
3602 DIFF_STATUS_UNMERGED,
3603 DIFF_STATUS_FILTER_AON,
3604 DIFF_STATUS_FILTER_BROKEN,
3605 '\0',
3606};
3607static unsigned int filter_bit['Z' + 1];
3609static void prepare_filter_bits(void)
3611{
3612 int i;
3613 if (!filter_bit[DIFF_STATUS_ADDED]) {
3615 for (i = 0; diff_status_letters[i]; i++)
3616 filter_bit[(int) diff_status_letters[i]] = (1 << i);
3617 }
3618}
3619static unsigned filter_bit_tst(char status, const struct diff_options *opt)
3621{
3622 return opt->filter & filter_bit[(int) status];
3623}
3624static int parse_diff_filter_opt(const char *optarg, struct diff_options *opt)
3626{
3627 int i, optch;
3628 prepare_filter_bits();
3630 /*
3632 * If there is a negation e.g. 'd' in the input, and we haven't
3633 * initialized the filter field with another --diff-filter, start
3634 * from full set of bits, except for AON.
3635 */
3636 if (!opt->filter) {
3637 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3638 if (optch < 'a' || 'z' < optch)
3639 continue;
3640 opt->filter = (1 << (ARRAY_SIZE(diff_status_letters) - 1)) - 1;
3641 opt->filter &= ~filter_bit[DIFF_STATUS_FILTER_AON];
3642 break;
3643 }
3644 }
3645 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3647 unsigned int bit;
3648 int negate;
3649 if ('a' <= optch && optch <= 'z') {
3651 negate = 1;
3652 optch = toupper(optch);
3653 } else {
3654 negate = 0;
3655 }
3656 bit = (0 <= optch && optch <= 'Z') ? filter_bit[optch] : 0;
3658 if (!bit)
3659 return optarg[i];
3660 if (negate)
3661 opt->filter &= ~bit;
3662 else
3663 opt->filter |= bit;
3664 }
3665 return 0;
3666}
3667static void enable_patch_output(int *fmt) {
3669 *fmt &= ~DIFF_FORMAT_NO_OUTPUT;
3670 *fmt |= DIFF_FORMAT_PATCH;
3671}
3672static int parse_one_token(const char **arg, const char *token)
3674{
3675 const char *rest;
3676 if (skip_prefix(*arg, token, &rest) && (!*rest || *rest == ',')) {
3677 *arg = rest;
3678 return 1;
3679 }
3680 return 0;
3681}
3682static int parse_ws_error_highlight(struct diff_options *opt, const char *arg)
3684{
3685 const char *orig_arg = arg;
3686 unsigned val = 0;
3687 while (*arg) {
3688 if (parse_one_token(&arg, "none"))
3689 val = 0;
3690 else if (parse_one_token(&arg, "default"))
3691 val = WSEH_NEW;
3692 else if (parse_one_token(&arg, "all"))
3693 val = WSEH_NEW | WSEH_OLD | WSEH_CONTEXT;
3694 else if (parse_one_token(&arg, "new"))
3695 val |= WSEH_NEW;
3696 else if (parse_one_token(&arg, "old"))
3697 val |= WSEH_OLD;
3698 else if (parse_one_token(&arg, "context"))
3699 val |= WSEH_CONTEXT;
3700 else {
3701 error("unknown value after ws-error-highlight=%.*s",
3702 (int)(arg - orig_arg), orig_arg);
3703 return 0;
3704 }
3705 if (*arg)
3706 arg++;
3707 }
3708 opt->ws_error_highlight = val;
3709 return 1;
3710}
3711int diff_opt_parse(struct diff_options *options,
3713 const char **av, int ac, const char *prefix)
3714{
3715 const char *arg = av[0];
3716 const char *optarg;
3717 int argcount;
3718 if (!prefix)
3720 prefix = "";
3721 /* Output format options */
3723 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch")
3724 || opt_arg(arg, 'U', "unified", &options->context))
3725 enable_patch_output(&options->output_format);
3726 else if (!strcmp(arg, "--raw"))
3727 options->output_format |= DIFF_FORMAT_RAW;
3728 else if (!strcmp(arg, "--patch-with-raw")) {
3729 enable_patch_output(&options->output_format);
3730 options->output_format |= DIFF_FORMAT_RAW;
3731 } else if (!strcmp(arg, "--numstat"))
3732 options->output_format |= DIFF_FORMAT_NUMSTAT;
3733 else if (!strcmp(arg, "--shortstat"))
3734 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3735 else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3736 return parse_dirstat_opt(options, "");
3737 else if (skip_prefix(arg, "-X", &arg))
3738 return parse_dirstat_opt(options, arg);
3739 else if (skip_prefix(arg, "--dirstat=", &arg))
3740 return parse_dirstat_opt(options, arg);
3741 else if (!strcmp(arg, "--cumulative"))
3742 return parse_dirstat_opt(options, "cumulative");
3743 else if (!strcmp(arg, "--dirstat-by-file"))
3744 return parse_dirstat_opt(options, "files");
3745 else if (skip_prefix(arg, "--dirstat-by-file=", &arg)) {
3746 parse_dirstat_opt(options, "files");
3747 return parse_dirstat_opt(options, arg);
3748 }
3749 else if (!strcmp(arg, "--check"))
3750 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3751 else if (!strcmp(arg, "--summary"))
3752 options->output_format |= DIFF_FORMAT_SUMMARY;
3753 else if (!strcmp(arg, "--patch-with-stat")) {
3754 enable_patch_output(&options->output_format);
3755 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3756 } else if (!strcmp(arg, "--name-only"))
3757 options->output_format |= DIFF_FORMAT_NAME;
3758 else if (!strcmp(arg, "--name-status"))
3759 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3760 else if (!strcmp(arg, "-s") || !strcmp(arg, "--no-patch"))
3761 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3762 else if (starts_with(arg, "--stat"))
3763 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3764 return stat_opt(options, av);
3765 /* renames options */
3767 else if (starts_with(arg, "-B") || starts_with(arg, "--break-rewrites=") ||
3768 !strcmp(arg, "--break-rewrites")) {
3769 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3770 return error("invalid argument to -B: %s", arg+2);
3771 }
3772 else if (starts_with(arg, "-M") || starts_with(arg, "--find-renames=") ||
3773 !strcmp(arg, "--find-renames")) {
3774 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3775 return error("invalid argument to -M: %s", arg+2);
3776 options->detect_rename = DIFF_DETECT_RENAME;
3777 }
3778 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3779 options->irreversible_delete = 1;
3780 }
3781 else if (starts_with(arg, "-C") || starts_with(arg, "--find-copies=") ||
3782 !strcmp(arg, "--find-copies")) {
3783 if (options->detect_rename == DIFF_DETECT_COPY)
3784 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3785 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3786 return error("invalid argument to -C: %s", arg+2);
3787 options->detect_rename = DIFF_DETECT_COPY;
3788 }
3789 else if (!strcmp(arg, "--no-renames"))
3790 options->detect_rename = 0;
3791 else if (!strcmp(arg, "--rename-empty"))
3792 DIFF_OPT_SET(options, RENAME_EMPTY);
3793 else if (!strcmp(arg, "--no-rename-empty"))
3794 DIFF_OPT_CLR(options, RENAME_EMPTY);
3795 else if (!strcmp(arg, "--relative"))
3796 DIFF_OPT_SET(options, RELATIVE_NAME);
3797 else if (skip_prefix(arg, "--relative=", &arg)) {
3798 DIFF_OPT_SET(options, RELATIVE_NAME);
3799 options->prefix = arg;
3800 }
3801 /* xdiff options */
3803 else if (!strcmp(arg, "--minimal"))
3804 DIFF_XDL_SET(options, NEED_MINIMAL);
3805 else if (!strcmp(arg, "--no-minimal"))
3806 DIFF_XDL_CLR(options, NEED_MINIMAL);
3807 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3808 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3809 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3810 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3811 else if (!strcmp(arg, "--ignore-space-at-eol"))
3812 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3813 else if (!strcmp(arg, "--ignore-blank-lines"))
3814 DIFF_XDL_SET(options, IGNORE_BLANK_LINES);
3815 else if (!strcmp(arg, "--compaction-heuristic"))
3816 DIFF_XDL_SET(options, COMPACTION_HEURISTIC);
3817 else if (!strcmp(arg, "--no-compaction-heuristic"))
3818 DIFF_XDL_CLR(options, COMPACTION_HEURISTIC);
3819 else if (!strcmp(arg, "--patience"))
3820 options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3821 else if (!strcmp(arg, "--histogram"))
3822 options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3823 else if ((argcount = parse_long_opt("diff-algorithm", av, &optarg))) {
3824 long value = parse_algorithm_value(optarg);
3825 if (value < 0)
3826 return error("option diff-algorithm accepts \"myers\", "
3827 "\"minimal\", \"patience\" and \"histogram\"");
3828 /* clear out previous settings */
3829 DIFF_XDL_CLR(options, NEED_MINIMAL);
3830 options->xdl_opts &= ~XDF_DIFF_ALGORITHM_MASK;
3831 options->xdl_opts |= value;
3832 return argcount;
3833 }
3834 /* flags options */
3836 else if (!strcmp(arg, "--binary")) {
3837 enable_patch_output(&options->output_format);
3838 DIFF_OPT_SET(options, BINARY);
3839 }
3840 else if (!strcmp(arg, "--full-index"))
3841 DIFF_OPT_SET(options, FULL_INDEX);
3842 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3843 DIFF_OPT_SET(options, TEXT);
3844 else if (!strcmp(arg, "-R"))
3845 DIFF_OPT_SET(options, REVERSE_DIFF);
3846 else if (!strcmp(arg, "--find-copies-harder"))
3847 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3848 else if (!strcmp(arg, "--follow"))
3849 DIFF_OPT_SET(options, FOLLOW_RENAMES);
3850 else if (!strcmp(arg, "--no-follow")) {
3851 DIFF_OPT_CLR(options, FOLLOW_RENAMES);
3852 DIFF_OPT_CLR(options, DEFAULT_FOLLOW_RENAMES);
3853 } else if (!strcmp(arg, "--color"))
3854 options->use_color = 1;
3855 else if (skip_prefix(arg, "--color=", &arg)) {
3856 int value = git_config_colorbool(NULL, arg);
3857 if (value < 0)
3858 return error("option `color' expects \"always\", \"auto\", or \"never\"");
3859 options->use_color = value;
3860 }
3861 else if (!strcmp(arg, "--no-color"))
3862 options->use_color = 0;
3863 else if (!strcmp(arg, "--color-words")) {
3864 options->use_color = 1;
3865 options->word_diff = DIFF_WORDS_COLOR;
3866 }
3867 else if (skip_prefix(arg, "--color-words=", &arg)) {
3868 options->use_color = 1;
3869 options->word_diff = DIFF_WORDS_COLOR;
3870 options->word_regex = arg;
3871 }
3872 else if (!strcmp(arg, "--word-diff")) {
3873 if (options->word_diff == DIFF_WORDS_NONE)
3874 options->word_diff = DIFF_WORDS_PLAIN;
3875 }
3876 else if (skip_prefix(arg, "--word-diff=", &arg)) {
3877 if (!strcmp(arg, "plain"))
3878 options->word_diff = DIFF_WORDS_PLAIN;
3879 else if (!strcmp(arg, "color")) {
3880 options->use_color = 1;
3881 options->word_diff = DIFF_WORDS_COLOR;
3882 }
3883 else if (!strcmp(arg, "porcelain"))
3884 options->word_diff = DIFF_WORDS_PORCELAIN;
3885 else if (!strcmp(arg, "none"))
3886 options->word_diff = DIFF_WORDS_NONE;
3887 else
3888 die("bad --word-diff argument: %s", arg);
3889 }
3890 else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3891 if (options->word_diff == DIFF_WORDS_NONE)
3892 options->word_diff = DIFF_WORDS_PLAIN;
3893 options->word_regex = optarg;
3894 return argcount;
3895 }
3896 else if (!strcmp(arg, "--exit-code"))
3897 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3898 else if (!strcmp(arg, "--quiet"))
3899 DIFF_OPT_SET(options, QUICK);
3900 else if (!strcmp(arg, "--ext-diff"))
3901 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3902 else if (!strcmp(arg, "--no-ext-diff"))
3903 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3904 else if (!strcmp(arg, "--textconv"))
3905 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3906 else if (!strcmp(arg, "--no-textconv"))
3907 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3908 else if (!strcmp(arg, "--ignore-submodules")) {
3909 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3910 handle_ignore_submodules_arg(options, "all");
3911 } else if (skip_prefix(arg, "--ignore-submodules=", &arg)) {
3912 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3913 handle_ignore_submodules_arg(options, arg);
3914 } else if (!strcmp(arg, "--submodule"))
3915 DIFF_OPT_SET(options, SUBMODULE_LOG);
3916 else if (skip_prefix(arg, "--submodule=", &arg))
3917 return parse_submodule_opt(options, arg);
3918 else if (skip_prefix(arg, "--ws-error-highlight=", &arg))
3919 return parse_ws_error_highlight(options, arg);
3920 /* misc options */
3922 else if (!strcmp(arg, "-z"))
3923 options->line_termination = 0;
3924 else if ((argcount = short_opt('l', av, &optarg))) {
3925 options->rename_limit = strtoul(optarg, NULL, 10);
3926 return argcount;
3927 }
3928 else if ((argcount = short_opt('S', av, &optarg))) {
3929 options->pickaxe = optarg;
3930 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3931 return argcount;
3932 } else if ((argcount = short_opt('G', av, &optarg))) {
3933 options->pickaxe = optarg;
3934 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3935 return argcount;
3936 }
3937 else if (!strcmp(arg, "--pickaxe-all"))
3938 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3939 else if (!strcmp(arg, "--pickaxe-regex"))
3940 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3941 else if ((argcount = short_opt('O', av, &optarg))) {
3942 const char *path = prefix_filename(prefix, strlen(prefix), optarg);
3943 options->orderfile = xstrdup(path);
3944 return argcount;
3945 }
3946 else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3947 int offending = parse_diff_filter_opt(optarg, options);
3948 if (offending)
3949 die("unknown change class '%c' in --diff-filter=%s",
3950 offending, optarg);
3951 return argcount;
3952 }
3953 else if (!strcmp(arg, "--abbrev"))
3954 options->abbrev = DEFAULT_ABBREV;
3955 else if (skip_prefix(arg, "--abbrev=", &arg)) {
3956 options->abbrev = strtoul(arg, NULL, 10);
3957 if (options->abbrev < MINIMUM_ABBREV)
3958 options->abbrev = MINIMUM_ABBREV;
3959 else if (40 < options->abbrev)
3960 options->abbrev = 40;
3961 }
3962 else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3963 options->a_prefix = optarg;
3964 return argcount;
3965 }
3966 else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
3967 options->b_prefix = optarg;
3968 return argcount;
3969 }
3970 else if (!strcmp(arg, "--no-prefix"))
3971 options->a_prefix = options->b_prefix = "";
3972 else if (opt_arg(arg, '\0', "inter-hunk-context",
3973 &options->interhunkcontext))
3974 ;
3975 else if (!strcmp(arg, "-W"))
3976 DIFF_OPT_SET(options, FUNCCONTEXT);
3977 else if (!strcmp(arg, "--function-context"))
3978 DIFF_OPT_SET(options, FUNCCONTEXT);
3979 else if (!strcmp(arg, "--no-function-context"))
3980 DIFF_OPT_CLR(options, FUNCCONTEXT);
3981 else if ((argcount = parse_long_opt("output", av, &optarg))) {
3982 const char *path = prefix_filename(prefix, strlen(prefix), optarg);
3983 options->file = fopen(path, "w");
3984 if (!options->file)
3985 die_errno("Could not open '%s'", path);
3986 options->close_file = 1;
3987 if (options->use_color != GIT_COLOR_ALWAYS)
3988 options->use_color = GIT_COLOR_NEVER;
3989 return argcount;
3990 } else
3991 return 0;
3992 return 1;
3993}
3994int parse_rename_score(const char **cp_p)
3996{
3997 unsigned long num, scale;
3998 int ch, dot;
3999 const char *cp = *cp_p;
4000 num = 0;
4002 scale = 1;
4003 dot = 0;
4004 for (;;) {
4005 ch = *cp;
4006 if ( !dot && ch == '.' ) {
4007 scale = 1;
4008 dot = 1;
4009 } else if ( ch == '%' ) {
4010 scale = dot ? scale*100 : 100;
4011 cp++; /* % is always at the end */
4012 break;
4013 } else if ( ch >= '0' && ch <= '9' ) {
4014 if ( scale < 100000 ) {
4015 scale *= 10;
4016 num = (num*10) + (ch-'0');
4017 }
4018 } else {
4019 break;
4020 }
4021 cp++;
4022 }
4023 *cp_p = cp;
4024 /* user says num divided by scale and we say internally that
4026 * is MAX_SCORE * num / scale.
4027 */
4028 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
4029}
4030static int diff_scoreopt_parse(const char *opt)
4032{
4033 int opt1, opt2, cmd;
4034 if (*opt++ != '-')
4036 return -1;
4037 cmd = *opt++;
4038 if (cmd == '-') {
4039 /* convert the long-form arguments into short-form versions */
4040 if (skip_prefix(opt, "break-rewrites", &opt)) {
4041 if (*opt == 0 || *opt++ == '=')
4042 cmd = 'B';
4043 } else if (skip_prefix(opt, "find-copies", &opt)) {
4044 if (*opt == 0 || *opt++ == '=')
4045 cmd = 'C';
4046 } else if (skip_prefix(opt, "find-renames", &opt)) {
4047 if (*opt == 0 || *opt++ == '=')
4048 cmd = 'M';
4049 }
4050 }
4051 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
4052 return -1; /* that is not a -M, -C, or -B option */
4053 opt1 = parse_rename_score(&opt);
4055 if (cmd != 'B')
4056 opt2 = 0;
4057 else {
4058 if (*opt == 0)
4059 opt2 = 0;
4060 else if (*opt != '/')
4061 return -1; /* we expect -B80/99 or -B80 */
4062 else {
4063 opt++;
4064 opt2 = parse_rename_score(&opt);
4065 }
4066 }
4067 if (*opt != 0)
4068 return -1;
4069 return opt1 | (opt2 << 16);
4070}
4071struct diff_queue_struct diff_queued_diff;
4073void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
4075{
4076 ALLOC_GROW(queue->queue, queue->nr + 1, queue->alloc);
4077 queue->queue[queue->nr++] = dp;
4078}
4079struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
4081 struct diff_filespec *one,
4082 struct diff_filespec *two)
4083{
4084 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
4085 dp->one = one;
4086 dp->two = two;
4087 if (queue)
4088 diff_q(queue, dp);
4089 return dp;
4090}
4091void diff_free_filepair(struct diff_filepair *p)
4093{
4094 free_filespec(p->one);
4095 free_filespec(p->two);
4096 free(p);
4097}
4098/* This is different from find_unique_abbrev() in that
4100 * it stuffs the result with dots for alignment.
4101 */
4102const char *diff_unique_abbrev(const unsigned char *sha1, int len)
4103{
4104 int abblen;
4105 const char *abbrev;
4106 if (len == 40)
4107 return sha1_to_hex(sha1);
4108 abbrev = find_unique_abbrev(sha1, len);
4110 abblen = strlen(abbrev);
4111 if (abblen < 37) {
4112 static char hex[41];
4113 if (len < abblen && abblen <= len + 2)
4114 xsnprintf(hex, sizeof(hex), "%s%.*s", abbrev, len+3-abblen, "..");
4115 else
4116 xsnprintf(hex, sizeof(hex), "%s...", abbrev);
4117 return hex;
4118 }
4119 return sha1_to_hex(sha1);
4120}
4121static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
4123{
4124 int line_termination = opt->line_termination;
4125 int inter_name_termination = line_termination ? '\t' : '\0';
4126 fprintf(opt->file, "%s", diff_line_prefix(opt));
4128 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
4129 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
4130 diff_unique_abbrev(p->one->oid.hash, opt->abbrev));
4131 fprintf(opt->file, "%s ",
4132 diff_unique_abbrev(p->two->oid.hash, opt->abbrev));
4133 }
4134 if (p->score) {
4135 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
4136 inter_name_termination);
4137 } else {
4138 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
4139 }
4140 if (p->status == DIFF_STATUS_COPIED ||
4142 p->status == DIFF_STATUS_RENAMED) {
4143 const char *name_a, *name_b;
4144 name_a = p->one->path;
4145 name_b = p->two->path;
4146 strip_prefix(opt->prefix_length, &name_a, &name_b);
4147 write_name_quoted(name_a, opt->file, inter_name_termination);
4148 write_name_quoted(name_b, opt->file, line_termination);
4149 } else {
4150 const char *name_a, *name_b;
4151 name_a = p->one->mode ? p->one->path : p->two->path;
4152 name_b = NULL;
4153 strip_prefix(opt->prefix_length, &name_a, &name_b);
4154 write_name_quoted(name_a, opt->file, line_termination);
4155 }
4156}
4157int diff_unmodified_pair(struct diff_filepair *p)
4159{
4160 /* This function is written stricter than necessary to support
4161 * the currently implemented transformers, but the idea is to
4162 * let transformers to produce diff_filepairs any way they want,
4163 * and filter and clean them up here before producing the output.
4164 */
4165 struct diff_filespec *one = p->one, *two = p->two;
4166 if (DIFF_PAIR_UNMERGED(p))
4168 return 0; /* unmerged is interesting */
4169 /* deletion, addition, mode or type change
4171 * and rename are all interesting.
4172 */
4173 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
4174 DIFF_PAIR_MODE_CHANGED(p) ||
4175 strcmp(one->path, two->path))
4176 return 0;
4177 /* both are valid and point at the same path. that is, we are
4179 * dealing with a change.
4180 */
4181 if (one->oid_valid && two->oid_valid &&
4182 !oidcmp(&one->oid, &two->oid) &&
4183 !one->dirty_submodule && !two->dirty_submodule)
4184 return 1; /* no change */
4185 if (!one->oid_valid && !two->oid_valid)
4186 return 1; /* both look at the same file on the filesystem. */
4187 return 0;
4188}
4189static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
4191{
4192 if (diff_unmodified_pair(p))
4193 return;
4194 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4196 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4197 return; /* no tree diffs in patch format */
4198 run_diff(p, o);
4200}
4201static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
4203 struct diffstat_t *diffstat)
4204{
4205 if (diff_unmodified_pair(p))
4206 return;
4207 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4209 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4210 return; /* no useful stat for tree diffs */
4211 run_diffstat(p, o, diffstat);
4213}
4214static void diff_flush_checkdiff(struct diff_filepair *p,
4216 struct diff_options *o)
4217{
4218 if (diff_unmodified_pair(p))
4219 return;
4220 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4222 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4223 return; /* nothing to check in tree diffs */
4224 run_checkdiff(p, o);
4226}
4227int diff_queue_is_empty(void)
4229{
4230 struct diff_queue_struct *q = &diff_queued_diff;
4231 int i;
4232 for (i = 0; i < q->nr; i++)
4233 if (!diff_unmodified_pair(q->queue[i]))
4234 return 0;
4235 return 1;
4236}
4237#if DIFF_DEBUG
4239void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
4240{
4241 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
4242 x, one ? one : "",
4243 s->path,
4244 DIFF_FILE_VALID(s) ? "valid" : "invalid",
4245 s->mode,
4246 s->oid_valid ? oid_to_hex(&s->oid) : "");
4247 fprintf(stderr, "queue[%d] %s size %lu\n",
4248 x, one ? one : "",
4249 s->size);
4250}
4251void diff_debug_filepair(const struct diff_filepair *p, int i)
4253{
4254 diff_debug_filespec(p->one, i, "one");
4255 diff_debug_filespec(p->two, i, "two");
4256 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4257 p->score, p->status ? p->status : '?',
4258 p->one->rename_used, p->broken_pair);
4259}
4260void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4262{
4263 int i;
4264 if (msg)
4265 fprintf(stderr, "%s\n", msg);
4266 fprintf(stderr, "q->nr = %d\n", q->nr);
4267 for (i = 0; i < q->nr; i++) {
4268 struct diff_filepair *p = q->queue[i];
4269 diff_debug_filepair(p, i);
4270 }
4271}
4272#endif
4273static void diff_resolve_rename_copy(void)
4275{
4276 int i;
4277 struct diff_filepair *p;
4278 struct diff_queue_struct *q = &diff_queued_diff;
4279 diff_debug_queue("resolve-rename-copy", q);
4281 for (i = 0; i < q->nr; i++) {
4283 p = q->queue[i];
4284 p->status = 0; /* undecided */
4285 if (DIFF_PAIR_UNMERGED(p))
4286 p->status = DIFF_STATUS_UNMERGED;
4287 else if (!DIFF_FILE_VALID(p->one))
4288 p->status = DIFF_STATUS_ADDED;
4289 else if (!DIFF_FILE_VALID(p->two))
4290 p->status = DIFF_STATUS_DELETED;
4291 else if (DIFF_PAIR_TYPE_CHANGED(p))
4292 p->status = DIFF_STATUS_TYPE_CHANGED;
4293 /* from this point on, we are dealing with a pair
4295 * whose both sides are valid and of the same type, i.e.
4296 * either in-place edit or rename/copy edit.
4297 */
4298 else if (DIFF_PAIR_RENAME(p)) {
4299 /*
4300 * A rename might have re-connected a broken
4301 * pair up, causing the pathnames to be the
4302 * same again. If so, that's not a rename at
4303 * all, just a modification..
4304 *
4305 * Otherwise, see if this source was used for
4306 * multiple renames, in which case we decrement
4307 * the count, and call it a copy.
4308 */
4309 if (!strcmp(p->one->path, p->two->path))
4310 p->status = DIFF_STATUS_MODIFIED;
4311 else if (--p->one->rename_used > 0)
4312 p->status = DIFF_STATUS_COPIED;
4313 else
4314 p->status = DIFF_STATUS_RENAMED;
4315 }
4316 else if (oidcmp(&p->one->oid, &p->two->oid) ||
4317 p->one->mode != p->two->mode ||
4318 p->one->dirty_submodule ||
4319 p->two->dirty_submodule ||
4320 is_null_oid(&p->one->oid))
4321 p->status = DIFF_STATUS_MODIFIED;
4322 else {
4323 /* This is a "no-change" entry and should not
4324 * happen anymore, but prepare for broken callers.
4325 */
4326 error("feeding unmodified %s to diffcore",
4327 p->one->path);
4328 p->status = DIFF_STATUS_UNKNOWN;
4329 }
4330 }
4331 diff_debug_queue("resolve-rename-copy done", q);
4332}
4333static int check_pair_status(struct diff_filepair *p)
4335{
4336 switch (p->status) {
4337 case DIFF_STATUS_UNKNOWN:
4338 return 0;
4339 case 0:
4340 die("internal error in diff-resolve-rename-copy");
4341 default:
4342 return 1;
4343 }
4344}
4345static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4347{
4348 int fmt = opt->output_format;
4349 if (fmt & DIFF_FORMAT_CHECKDIFF)
4351 diff_flush_checkdiff(p, opt);
4352 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4353 diff_flush_raw(p, opt);
4354 else if (fmt & DIFF_FORMAT_NAME) {
4355 const char *name_a, *name_b;
4356 name_a = p->two->path;
4357 name_b = NULL;
4358 strip_prefix(opt->prefix_length, &name_a, &name_b);
4359 write_name_quoted(name_a, opt->file, opt->line_termination);
4360 }
4361}
4362static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4364{
4365 if (fs->mode)
4366 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4367 else
4368 fprintf(file, " %s ", newdelete);
4369 write_name_quoted(fs->path, file, '\n');
4370}
4371static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4374 const char *line_prefix)
4375{
4376 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4377 fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4378 p->two->mode, show_name ? ' ' : '\n');
4379 if (show_name) {
4380 write_name_quoted(p->two->path, file, '\n');
4381 }
4382 }
4383}
4384static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4386 const char *line_prefix)
4387{
4388 char *names = pprint_rename(p->one->path, p->two->path);
4389 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4391 free(names);
4392 show_mode_change(file, p, 0, line_prefix);
4393}
4394static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4396{
4397 FILE *file = opt->file;
4398 const char *line_prefix = diff_line_prefix(opt);
4399 switch(p->status) {
4401 case DIFF_STATUS_DELETED:
4402 fputs(line_prefix, file);
4403 show_file_mode_name(file, "delete", p->one);
4404 break;
4405 case DIFF_STATUS_ADDED:
4406 fputs(line_prefix, file);
4407 show_file_mode_name(file, "create", p->two);
4408 break;
4409 case DIFF_STATUS_COPIED:
4410 fputs(line_prefix, file);
4411 show_rename_copy(file, "copy", p, line_prefix);
4412 break;
4413 case DIFF_STATUS_RENAMED:
4414 fputs(line_prefix, file);
4415 show_rename_copy(file, "rename", p, line_prefix);
4416 break;
4417 default:
4418 if (p->score) {
4419 fprintf(file, "%s rewrite ", line_prefix);
4420 write_name_quoted(p->two->path, file, ' ');
4421 fprintf(file, "(%d%%)\n", similarity_index(p));
4422 }
4423 show_mode_change(file, p, !p->score, line_prefix);
4424 break;
4425 }
4426}
4427struct patch_id_t {
4429 git_SHA_CTX *ctx;
4430 int patchlen;
4431};
4432static int remove_space(char *line, int len)
4434{
4435 int i;
4436 char *dst = line;
4437 unsigned char c;
4438 for (i = 0; i < len; i++)
4440 if (!isspace((c = line[i])))
4441 *dst++ = c;
4442 return dst - line;
4444}
4445static void patch_id_consume(void *priv, char *line, unsigned long len)
4447{
4448 struct patch_id_t *data = priv;
4449 int new_len;
4450 /* Ignore line numbers when computing the SHA1 of the patch */
4452 if (starts_with(line, "@@ -"))
4453 return;
4454 new_len = remove_space(line, len);
4456 git_SHA1_Update(data->ctx, line, new_len);
4458 data->patchlen += new_len;
4459}
4460/* returns 0 upon success, and writes result into sha1 */
4462static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1, int diff_header_only)
4463{
4464 struct diff_queue_struct *q = &diff_queued_diff;
4465 int i;
4466 git_SHA_CTX ctx;
4467 struct patch_id_t data;
4468 char buffer[PATH_MAX * 4 + 20];
4469 git_SHA1_Init(&ctx);
4471 memset(&data, 0, sizeof(struct patch_id_t));
4472 data.ctx = &ctx;
4473 for (i = 0; i < q->nr; i++) {
4475 xpparam_t xpp;
4476 xdemitconf_t xecfg;
4477 mmfile_t mf1, mf2;
4478 struct diff_filepair *p = q->queue[i];
4479 int len1, len2;
4480 memset(&xpp, 0, sizeof(xpp));
4482 memset(&xecfg, 0, sizeof(xecfg));
4483 if (p->status == 0)
4484 return error("internal diff status error");
4485 if (p->status == DIFF_STATUS_UNKNOWN)
4486 continue;
4487 if (diff_unmodified_pair(p))
4488 continue;
4489 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4490 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4491 continue;
4492 if (DIFF_PAIR_UNMERGED(p))
4493 continue;
4494 diff_fill_sha1_info(p->one);
4496 diff_fill_sha1_info(p->two);
4497 len1 = remove_space(p->one->path, strlen(p->one->path));
4499 len2 = remove_space(p->two->path, strlen(p->two->path));
4500 if (p->one->mode == 0)
4501 len1 = snprintf(buffer, sizeof(buffer),
4502 "diff--gita/%.*sb/%.*s"
4503 "newfilemode%06o"
4504 "---/dev/null"
4505 "+++b/%.*s",
4506 len1, p->one->path,
4507 len2, p->two->path,
4508 p->two->mode,
4509 len2, p->two->path);
4510 else if (p->two->mode == 0)
4511 len1 = snprintf(buffer, sizeof(buffer),
4512 "diff--gita/%.*sb/%.*s"
4513 "deletedfilemode%06o"
4514 "---a/%.*s"
4515 "+++/dev/null",
4516 len1, p->one->path,
4517 len2, p->two->path,
4518 p->one->mode,
4519 len1, p->one->path);
4520 else
4521 len1 = snprintf(buffer, sizeof(buffer),
4522 "diff--gita/%.*sb/%.*s"
4523 "---a/%.*s"
4524 "+++b/%.*s",
4525 len1, p->one->path,
4526 len2, p->two->path,
4527 len1, p->one->path,
4528 len2, p->two->path);
4529 git_SHA1_Update(&ctx, buffer, len1);
4530 if (diff_header_only)
4532 continue;
4533 if (fill_mmfile(&mf1, p->one) < 0 ||
4535 fill_mmfile(&mf2, p->two) < 0)
4536 return error("unable to read files to diff");
4537 if (diff_filespec_is_binary(p->one) ||
4539 diff_filespec_is_binary(p->two)) {
4540 git_SHA1_Update(&ctx, oid_to_hex(&p->one->oid),
4541 40);
4542 git_SHA1_Update(&ctx, oid_to_hex(&p->two->oid),
4543 40);
4544 continue;
4545 }
4546 xpp.flags = 0;
4548 xecfg.ctxlen = 3;
4549 xecfg.flags = 0;
4550 if (xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4551 &xpp, &xecfg))
4552 return error("unable to generate patch-id diff for %s",
4553 p->one->path);
4554 }
4555 git_SHA1_Final(sha1, &ctx);
4557 return 0;
4558}
4559int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1, int diff_header_only)
4561{
4562 struct diff_queue_struct *q = &diff_queued_diff;
4563 int i;
4564 int result = diff_get_patch_id(options, sha1, diff_header_only);
4565 for (i = 0; i < q->nr; i++)
4567 diff_free_filepair(q->queue[i]);
4568 free(q->queue);
4570 DIFF_QUEUE_CLEAR(q);
4571 return result;
4573}
4574static int is_summary_empty(const struct diff_queue_struct *q)
4576{
4577 int i;
4578 for (i = 0; i < q->nr; i++) {
4580 const struct diff_filepair *p = q->queue[i];
4581 switch (p->status) {
4583 case DIFF_STATUS_DELETED:
4584 case DIFF_STATUS_ADDED:
4585 case DIFF_STATUS_COPIED:
4586 case DIFF_STATUS_RENAMED:
4587 return 0;
4588 default:
4589 if (p->score)
4590 return 0;
4591 if (p->one->mode && p->two->mode &&
4592 p->one->mode != p->two->mode)
4593 return 0;
4594 break;
4595 }
4596 }
4597 return 1;
4598}
4599static const char rename_limit_warning[] =
4601"inexact rename detection was skipped due to too many files.";
4602static const char degrade_cc_to_c_warning[] =
4604"only found copies from modified paths due to too many files.";
4605static const char rename_limit_advice[] =
4607"you may want to set your %s variable to at least "
4608"%d and retry the command.";
4609void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4611{
4612 if (degraded_cc)
4613 warning(degrade_cc_to_c_warning);
4614 else if (needed)
4615 warning(rename_limit_warning);
4616 else
4617 return;
4618 if (0 < needed && needed < 32767)
4619 warning(rename_limit_advice, varname, needed);
4620}
4621void diff_flush(struct diff_options *options)
4623{
4624 struct diff_queue_struct *q = &diff_queued_diff;
4625 int i, output_format = options->output_format;
4626 int separator = 0;
4627 int dirstat_by_line = 0;
4628 /*
4630 * Order: raw, stat, summary, patch
4631 * or: name/name-status/checkdiff (other bits clear)
4632 */
4633 if (!q->nr)
4634 goto free_queue;
4635 if (output_format & (DIFF_FORMAT_RAW |
4637 DIFF_FORMAT_NAME |
4638 DIFF_FORMAT_NAME_STATUS |
4639 DIFF_FORMAT_CHECKDIFF)) {
4640 for (i = 0; i < q->nr; i++) {
4641 struct diff_filepair *p = q->queue[i];
4642 if (check_pair_status(p))
4643 flush_one_pair(p, options);
4644 }
4645 separator++;
4646 }
4647 if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4649 dirstat_by_line = 1;
4650 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4652 dirstat_by_line) {
4653 struct diffstat_t diffstat;
4654 memset(&diffstat, 0, sizeof(struct diffstat_t));
4656 for (i = 0; i < q->nr; i++) {
4657 struct diff_filepair *p = q->queue[i];
4658 if (check_pair_status(p))
4659 diff_flush_stat(p, options, &diffstat);
4660 }
4661 if (output_format & DIFF_FORMAT_NUMSTAT)
4662 show_numstat(&diffstat, options);
4663 if (output_format & DIFF_FORMAT_DIFFSTAT)
4664 show_stats(&diffstat, options);
4665 if (output_format & DIFF_FORMAT_SHORTSTAT)
4666 show_shortstats(&diffstat, options);
4667 if (output_format & DIFF_FORMAT_DIRSTAT && dirstat_by_line)
4668 show_dirstat_by_line(&diffstat, options);
4669 free_diffstat_info(&diffstat);
4670 separator++;
4671 }
4672 if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4673 show_dirstat(options);
4674 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4676 for (i = 0; i < q->nr; i++) {
4677 diff_summary(options, q->queue[i]);
4678 }
4679 separator++;
4680 }
4681 if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4683 DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4684 DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4685 /*
4686 * run diff_flush_patch for the exit status. setting
4687 * options->file to /dev/null should be safe, because we
4688 * aren't supposed to produce any output anyway.
4689 */
4690 if (options->close_file)
4691 fclose(options->file);
4692 options->file = fopen("/dev/null", "w");
4693 if (!options->file)
4694 die_errno("Could not open /dev/null");
4695 options->close_file = 1;
4696 for (i = 0; i < q->nr; i++) {
4697 struct diff_filepair *p = q->queue[i];
4698 if (check_pair_status(p))
4699 diff_flush_patch(p, options);
4700 if (options->found_changes)
4701 break;
4702 }
4703 }
4704 if (output_format & DIFF_FORMAT_PATCH) {
4706 if (separator) {
4707 fprintf(options->file, "%s%c",
4708 diff_line_prefix(options),
4709 options->line_termination);
4710 if (options->stat_sep) {
4711 /* attach patch instead of inline */
4712 fputs(options->stat_sep, options->file);
4713 }
4714 }
4715 for (i = 0; i < q->nr; i++) {
4717 struct diff_filepair *p = q->queue[i];
4718 if (check_pair_status(p))
4719 diff_flush_patch(p, options);
4720 }
4721 }
4722 if (output_format & DIFF_FORMAT_CALLBACK)
4724 options->format_callback(q, options, options->format_callback_data);
4725 for (i = 0; i < q->nr; i++)
4727 diff_free_filepair(q->queue[i]);
4728free_queue:
4729 free(q->queue);
4730 DIFF_QUEUE_CLEAR(q);
4731 if (options->close_file)
4732 fclose(options->file);
4733 /*
4735 * Report the content-level differences with HAS_CHANGES;
4736 * diff_addremove/diff_change does not set the bit when
4737 * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4738 */
4739 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4740 if (options->found_changes)
4741 DIFF_OPT_SET(options, HAS_CHANGES);
4742 else
4743 DIFF_OPT_CLR(options, HAS_CHANGES);
4744 }
4745}
4746static int match_filter(const struct diff_options *options, const struct diff_filepair *p)
4748{
4749 return (((p->status == DIFF_STATUS_MODIFIED) &&
4750 ((p->score &&
4751 filter_bit_tst(DIFF_STATUS_FILTER_BROKEN, options)) ||
4752 (!p->score &&
4753 filter_bit_tst(DIFF_STATUS_MODIFIED, options)))) ||
4754 ((p->status != DIFF_STATUS_MODIFIED) &&
4755 filter_bit_tst(p->status, options)));
4756}
4757static void diffcore_apply_filter(struct diff_options *options)
4759{
4760 int i;
4761 struct diff_queue_struct *q = &diff_queued_diff;
4762 struct diff_queue_struct outq;
4763 DIFF_QUEUE_CLEAR(&outq);
4765 if (!options->filter)
4767 return;
4768 if (filter_bit_tst(DIFF_STATUS_FILTER_AON, options)) {
4770 int found;
4771 for (i = found = 0; !found && i < q->nr; i++) {
4772 if (match_filter(options, q->queue[i]))
4773 found++;
4774 }
4775 if (found)
4776 return;
4777 /* otherwise we will clear the whole queue
4779 * by copying the empty outq at the end of this
4780 * function, but first clear the current entries
4781 * in the queue.
4782 */
4783 for (i = 0; i < q->nr; i++)
4784 diff_free_filepair(q->queue[i]);
4785 }
4786 else {
4787 /* Only the matching ones */
4788 for (i = 0; i < q->nr; i++) {
4789 struct diff_filepair *p = q->queue[i];
4790 if (match_filter(options, p))
4791 diff_q(&outq, p);
4792 else
4793 diff_free_filepair(p);
4794 }
4795 }
4796 free(q->queue);
4797 *q = outq;
4798}
4799/* Check whether two filespecs with the same mode and size are identical */
4801static int diff_filespec_is_identical(struct diff_filespec *one,
4802 struct diff_filespec *two)
4803{
4804 if (S_ISGITLINK(one->mode))
4805 return 0;
4806 if (diff_populate_filespec(one, 0))
4807 return 0;
4808 if (diff_populate_filespec(two, 0))
4809 return 0;
4810 return !memcmp(one->data, two->data, one->size);
4811}
4812static int diff_filespec_check_stat_unmatch(struct diff_filepair *p)
4814{
4815 if (p->done_skip_stat_unmatch)
4816 return p->skip_stat_unmatch_result;
4817 p->done_skip_stat_unmatch = 1;
4819 p->skip_stat_unmatch_result = 0;
4820 /*
4821 * 1. Entries that come from stat info dirtiness
4822 * always have both sides (iow, not create/delete),
4823 * one side of the object name is unknown, with
4824 * the same mode and size. Keep the ones that
4825 * do not match these criteria. They have real
4826 * differences.
4827 *
4828 * 2. At this point, the file is known to be modified,
4829 * with the same mode and size, and the object
4830 * name of one side is unknown. Need to inspect
4831 * the identical contents.
4832 */
4833 if (!DIFF_FILE_VALID(p->one) || /* (1) */
4834 !DIFF_FILE_VALID(p->two) ||
4835 (p->one->oid_valid && p->two->oid_valid) ||
4836 (p->one->mode != p->two->mode) ||
4837 diff_populate_filespec(p->one, CHECK_SIZE_ONLY) ||
4838 diff_populate_filespec(p->two, CHECK_SIZE_ONLY) ||
4839 (p->one->size != p->two->size) ||
4840 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4841 p->skip_stat_unmatch_result = 1;
4842 return p->skip_stat_unmatch_result;
4843}
4844static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4846{
4847 int i;
4848 struct diff_queue_struct *q = &diff_queued_diff;
4849 struct diff_queue_struct outq;
4850 DIFF_QUEUE_CLEAR(&outq);
4851 for (i = 0; i < q->nr; i++) {
4853 struct diff_filepair *p = q->queue[i];
4854 if (diff_filespec_check_stat_unmatch(p))
4856 diff_q(&outq, p);
4857 else {
4858 /*
4859 * The caller can subtract 1 from skip_stat_unmatch
4860 * to determine how many paths were dirty only
4861 * due to stat info mismatch.
4862 */
4863 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4864 diffopt->skip_stat_unmatch++;
4865 diff_free_filepair(p);
4866 }
4867 }
4868 free(q->queue);
4869 *q = outq;
4870}
4871static int diffnamecmp(const void *a_, const void *b_)
4873{
4874 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4875 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4876 const char *name_a, *name_b;
4877 name_a = a->one ? a->one->path : a->two->path;
4879 name_b = b->one ? b->one->path : b->two->path;
4880 return strcmp(name_a, name_b);
4881}
4882void diffcore_fix_diff_index(struct diff_options *options)
4884{
4885 struct diff_queue_struct *q = &diff_queued_diff;
4886 qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4887}
4888void diffcore_std(struct diff_options *options)
4890{
4891 /* NOTE please keep the following in sync with diff_tree_combined() */
4892 if (options->skip_stat_unmatch)
4893 diffcore_skip_stat_unmatch(options);
4894 if (!options->found_follow) {
4895 /* See try_to_follow_renames() in tree-diff.c */
4896 if (options->break_opt != -1)
4897 diffcore_break(options->break_opt);
4898 if (options->detect_rename)
4899 diffcore_rename(options);
4900 if (options->break_opt != -1)
4901 diffcore_merge_broken();
4902 }
4903 if (options->pickaxe)
4904 diffcore_pickaxe(options);
4905 if (options->orderfile)
4906 diffcore_order(options->orderfile);
4907 if (!options->found_follow)
4908 /* See try_to_follow_renames() in tree-diff.c */
4909 diff_resolve_rename_copy();
4910 diffcore_apply_filter(options);
4911 if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4913 DIFF_OPT_SET(options, HAS_CHANGES);
4914 else
4915 DIFF_OPT_CLR(options, HAS_CHANGES);
4916 options->found_follow = 0;
4918}
4919int diff_result_code(struct diff_options *opt, int status)
4921{
4922 int result = 0;
4923 diff_warn_rename_limit("diff.renameLimit",
4925 opt->needed_rename_limit,
4926 opt->degraded_cc_to_c);
4927 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4928 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4929 return status;
4930 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4931 DIFF_OPT_TST(opt, HAS_CHANGES))
4932 result |= 01;
4933 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4934 DIFF_OPT_TST(opt, CHECK_FAILED))
4935 result |= 02;
4936 return result;
4937}
4938int diff_can_quit_early(struct diff_options *opt)
4940{
4941 return (DIFF_OPT_TST(opt, QUICK) &&
4942 !opt->filter &&
4943 DIFF_OPT_TST(opt, HAS_CHANGES));
4944}
4945/*
4947 * Shall changes to this submodule be ignored?
4948 *
4949 * Submodule changes can be configured to be ignored separately for each path,
4950 * but that configuration can be overridden from the command line.
4951 */
4952static int is_submodule_ignored(const char *path, struct diff_options *options)
4953{
4954 int ignored = 0;
4955 unsigned orig_flags = options->flags;
4956 if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4957 set_diffopt_flags_from_submodule_config(options, path);
4958 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4959 ignored = 1;
4960 options->flags = orig_flags;
4961 return ignored;
4962}
4963void diff_addremove(struct diff_options *options,
4965 int addremove, unsigned mode,
4966 const unsigned char *sha1,
4967 int sha1_valid,
4968 const char *concatpath, unsigned dirty_submodule)
4969{
4970 struct diff_filespec *one, *two;
4971 if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
4973 return;
4974 /* This may look odd, but it is a preparation for
4976 * feeding "there are unchanged files which should
4977 * not produce diffs, but when you are doing copy
4978 * detection you would need them, so here they are"
4979 * entries to the diff-core. They will be prefixed
4980 * with something like '=' or '*' (I haven't decided
4981 * which but should not make any difference).
4982 * Feeding the same new and old to diff_change()
4983 * also has the same effect.
4984 * Before the final output happens, they are pruned after
4985 * merged into rename/copy pairs as appropriate.
4986 */
4987 if (DIFF_OPT_TST(options, REVERSE_DIFF))
4988 addremove = (addremove == '+' ? '-' :
4989 addremove == '-' ? '+' : addremove);
4990 if (options->prefix &&
4992 strncmp(concatpath, options->prefix, options->prefix_length))
4993 return;
4994 one = alloc_filespec(concatpath);
4996 two = alloc_filespec(concatpath);
4997 if (addremove != '+')
4999 fill_filespec(one, sha1, sha1_valid, mode);
5000 if (addremove != '-') {
5001 fill_filespec(two, sha1, sha1_valid, mode);
5002 two->dirty_submodule = dirty_submodule;
5003 }
5004 diff_queue(&diff_queued_diff, one, two);
5006 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5007 DIFF_OPT_SET(options, HAS_CHANGES);
5008}
5009void diff_change(struct diff_options *options,
5011 unsigned old_mode, unsigned new_mode,
5012 const unsigned char *old_sha1,
5013 const unsigned char *new_sha1,
5014 int old_sha1_valid, int new_sha1_valid,
5015 const char *concatpath,
5016 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
5017{
5018 struct diff_filespec *one, *two;
5019 struct diff_filepair *p;
5020 if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
5022 is_submodule_ignored(concatpath, options))
5023 return;
5024 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
5026 unsigned tmp;
5027 const unsigned char *tmp_c;
5028 tmp = old_mode; old_mode = new_mode; new_mode = tmp;
5029 tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
5030 tmp = old_sha1_valid; old_sha1_valid = new_sha1_valid;
5031 new_sha1_valid = tmp;
5032 tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
5033 new_dirty_submodule = tmp;
5034 }
5035 if (options->prefix &&
5037 strncmp(concatpath, options->prefix, options->prefix_length))
5038 return;
5039 one = alloc_filespec(concatpath);
5041 two = alloc_filespec(concatpath);
5042 fill_filespec(one, old_sha1, old_sha1_valid, old_mode);
5043 fill_filespec(two, new_sha1, new_sha1_valid, new_mode);
5044 one->dirty_submodule = old_dirty_submodule;
5045 two->dirty_submodule = new_dirty_submodule;
5046 p = diff_queue(&diff_queued_diff, one, two);
5047 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5049 return;
5050 if (DIFF_OPT_TST(options, QUICK) && options->skip_stat_unmatch &&
5052 !diff_filespec_check_stat_unmatch(p))
5053 return;
5054 DIFF_OPT_SET(options, HAS_CHANGES);
5056}
5057struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
5059{
5060 struct diff_filepair *pair;
5061 struct diff_filespec *one, *two;
5062 if (options->prefix &&
5064 strncmp(path, options->prefix, options->prefix_length))
5065 return NULL;
5066 one = alloc_filespec(path);
5068 two = alloc_filespec(path);
5069 pair = diff_queue(&diff_queued_diff, one, two);
5070 pair->is_unmerged = 1;
5071 return pair;
5072}
5073static char *run_textconv(const char *pgm, struct diff_filespec *spec,
5075 size_t *outsize)
5076{
5077 struct diff_tempfile *temp;
5078 const char *argv[3];
5079 const char **arg = argv;
5080 struct child_process child = CHILD_PROCESS_INIT;
5081 struct strbuf buf = STRBUF_INIT;
5082 int err = 0;
5083 temp = prepare_temp_file(spec->path, spec);
5085 *arg++ = pgm;
5086 *arg++ = temp->name;
5087 *arg = NULL;
5088 child.use_shell = 1;
5090 child.argv = argv;
5091 child.out = -1;
5092 if (start_command(&child)) {
5093 remove_tempfile();
5094 return NULL;
5095 }
5096 if (strbuf_read(&buf, child.out, 0) < 0)
5098 err = error("error reading from textconv command '%s'", pgm);
5099 close(child.out);
5100 if (finish_command(&child) || err) {
5102 strbuf_release(&buf);
5103 remove_tempfile();
5104 return NULL;
5105 }
5106 remove_tempfile();
5107 return strbuf_detach(&buf, outsize);
5109}
5110size_t fill_textconv(struct userdiff_driver *driver,
5112 struct diff_filespec *df,
5113 char **outbuf)
5114{
5115 size_t size;
5116 if (!driver) {
5118 if (!DIFF_FILE_VALID(df)) {
5119 *outbuf = "";
5120 return 0;
5121 }
5122 if (diff_populate_filespec(df, 0))
5123 die("unable to read files to diff");
5124 *outbuf = df->data;
5125 return df->size;
5126 }
5127 if (!driver->textconv)
5129 die("BUG: fill_textconv called with non-textconv driver");
5130 if (driver->textconv_cache && df->oid_valid) {
5132 *outbuf = notes_cache_get(driver->textconv_cache,
5133 df->oid.hash,
5134 &size);
5135 if (*outbuf)
5136 return size;
5137 }
5138 *outbuf = run_textconv(driver->textconv, df, &size);
5140 if (!*outbuf)
5141 die("unable to read files to diff");
5142 if (driver->textconv_cache && df->oid_valid) {
5144 /* ignore errors, as we might be in a readonly repository */
5145 notes_cache_put(driver->textconv_cache, df->oid.hash, *outbuf,
5146 size);
5147 /*
5148 * we could save up changes and flush them all at the end,
5149 * but we would need an extra call after all diffing is done.
5150 * Since generating a cache entry is the slow path anyway,
5151 * this extra overhead probably isn't a big deal.
5152 */
5153 notes_cache_write(driver->textconv_cache);
5154 }
5155 return size;
5157}
5158void setup_diff_pager(struct diff_options *opt)
5160{
5161 /*
5162 * If the user asked for our exit code, then either they want --quiet
5163 * or --exit-code. We should definitely not bother with a pager in the
5164 * former case, as we will generate no output. Since we still properly
5165 * report our exit code even when a pager is run, we _could_ run a
5166 * pager with --exit-code. But since we have not done so historically,
5167 * and because it is easy to find people oneline advising "git diff
5168 * --exit-code" in hooks and other scripts, we do not do so.
5169 */
5170 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5171 check_pager_config("diff") != 0)
5172 setup_pager();
5173}