1/*
2 * Copyright (C) 2005 Junio C Hamano
3 */
4#include "cache.h"
5#include "quote.h"
6#include "diff.h"
7#include "diffcore.h"
8#include "delta.h"
9#include "xdiff-interface.h"
10#include "color.h"
11#include "attr.h"
12#include "run-command.h"
13#include "utf8.h"
14#include "userdiff.h"
15#include "sigchain.h"
16#include "submodule.h"
17#include "ll-merge.h"
18
19#ifdef NO_FAST_WORKING_DIRECTORY
20#define FAST_WORKING_DIRECTORY 0
21#else
22#define FAST_WORKING_DIRECTORY 1
23#endif
24
25static int diff_detect_rename_default;
26static int diff_rename_limit_default = 400;
27static int diff_suppress_blank_empty;
28int diff_use_color_default = -1;
29static const char *diff_word_regex_cfg;
30static const char *external_diff_cmd_cfg;
31int diff_auto_refresh_index = 1;
32static int diff_mnemonic_prefix;
33static int diff_no_prefix;
34static int diff_stat_graph_width;
35static int diff_dirstat_permille_default = 30;
36static struct diff_options default_diff_options;
37
38static char diff_colors[][COLOR_MAXLEN] = {
39 GIT_COLOR_RESET,
40 GIT_COLOR_NORMAL, /* PLAIN */
41 GIT_COLOR_BOLD, /* METAINFO */
42 GIT_COLOR_CYAN, /* FRAGINFO */
43 GIT_COLOR_RED, /* OLD */
44 GIT_COLOR_GREEN, /* NEW */
45 GIT_COLOR_YELLOW, /* COMMIT */
46 GIT_COLOR_BG_RED, /* WHITESPACE */
47 GIT_COLOR_NORMAL, /* FUNCINFO */
48};
49
50static int parse_diff_color_slot(const char *var, int ofs)
51{
52 if (!strcasecmp(var+ofs, "plain"))
53 return DIFF_PLAIN;
54 if (!strcasecmp(var+ofs, "meta"))
55 return DIFF_METAINFO;
56 if (!strcasecmp(var+ofs, "frag"))
57 return DIFF_FRAGINFO;
58 if (!strcasecmp(var+ofs, "old"))
59 return DIFF_FILE_OLD;
60 if (!strcasecmp(var+ofs, "new"))
61 return DIFF_FILE_NEW;
62 if (!strcasecmp(var+ofs, "commit"))
63 return DIFF_COMMIT;
64 if (!strcasecmp(var+ofs, "whitespace"))
65 return DIFF_WHITESPACE;
66 if (!strcasecmp(var+ofs, "func"))
67 return DIFF_FUNCINFO;
68 return -1;
69}
70
71static int parse_dirstat_params(struct diff_options *options, const char *params,
72 struct strbuf *errmsg)
73{
74 const char *p = params;
75 int p_len, ret = 0;
76
77 while (*p) {
78 p_len = strchrnul(p, ',') - p;
79 if (!memcmp(p, "changes", p_len)) {
80 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
81 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
82 } else if (!memcmp(p, "lines", p_len)) {
83 DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
84 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
85 } else if (!memcmp(p, "files", p_len)) {
86 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
87 DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
88 } else if (!memcmp(p, "noncumulative", p_len)) {
89 DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
90 } else if (!memcmp(p, "cumulative", p_len)) {
91 DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
92 } else if (isdigit(*p)) {
93 char *end;
94 int permille = strtoul(p, &end, 10) * 10;
95 if (*end == '.' && isdigit(*++end)) {
96 /* only use first digit */
97 permille += *end - '0';
98 /* .. and ignore any further digits */
99 while (isdigit(*++end))
100 ; /* nothing */
101 }
102 if (end - p == p_len)
103 options->dirstat_permille = permille;
104 else {
105 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%.*s'\n"),
106 p_len, p);
107 ret++;
108 }
109 } else {
110 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%.*s'\n"),
111 p_len, p);
112 ret++;
113 }
114
115 p += p_len;
116
117 if (*p)
118 p++; /* more parameters, swallow separator */
119 }
120 return ret;
121}
122
123static int git_config_rename(const char *var, const char *value)
124{
125 if (!value)
126 return DIFF_DETECT_RENAME;
127 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
128 return DIFF_DETECT_COPY;
129 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
130}
131
132/*
133 * These are to give UI layer defaults.
134 * The core-level commands such as git-diff-files should
135 * never be affected by the setting of diff.renames
136 * the user happens to have in the configuration file.
137 */
138int git_diff_ui_config(const char *var, const char *value, void *cb)
139{
140 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
141 diff_use_color_default = git_config_colorbool(var, value);
142 return 0;
143 }
144 if (!strcmp(var, "diff.renames")) {
145 diff_detect_rename_default = git_config_rename(var, value);
146 return 0;
147 }
148 if (!strcmp(var, "diff.autorefreshindex")) {
149 diff_auto_refresh_index = git_config_bool(var, value);
150 return 0;
151 }
152 if (!strcmp(var, "diff.mnemonicprefix")) {
153 diff_mnemonic_prefix = git_config_bool(var, value);
154 return 0;
155 }
156 if (!strcmp(var, "diff.noprefix")) {
157 diff_no_prefix = git_config_bool(var, value);
158 return 0;
159 }
160 if (!strcmp(var, "diff.statgraphwidth")) {
161 diff_stat_graph_width = git_config_int(var, value);
162 return 0;
163 }
164 if (!strcmp(var, "diff.external"))
165 return git_config_string(&external_diff_cmd_cfg, var, value);
166 if (!strcmp(var, "diff.wordregex"))
167 return git_config_string(&diff_word_regex_cfg, var, value);
168
169 if (!strcmp(var, "diff.ignoresubmodules"))
170 handle_ignore_submodules_arg(&default_diff_options, value);
171
172 if (git_color_config(var, value, cb) < 0)
173 return -1;
174
175 return git_diff_basic_config(var, value, cb);
176}
177
178int git_diff_basic_config(const char *var, const char *value, void *cb)
179{
180 if (!strcmp(var, "diff.renamelimit")) {
181 diff_rename_limit_default = git_config_int(var, value);
182 return 0;
183 }
184
185 if (userdiff_config(var, value) < 0)
186 return -1;
187
188 if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
189 int slot = parse_diff_color_slot(var, 11);
190 if (slot < 0)
191 return 0;
192 if (!value)
193 return config_error_nonbool(var);
194 color_parse(value, var, diff_colors[slot]);
195 return 0;
196 }
197
198 /* like GNU diff's --suppress-blank-empty option */
199 if (!strcmp(var, "diff.suppressblankempty") ||
200 /* for backwards compatibility */
201 !strcmp(var, "diff.suppress-blank-empty")) {
202 diff_suppress_blank_empty = git_config_bool(var, value);
203 return 0;
204 }
205
206 if (!strcmp(var, "diff.dirstat")) {
207 struct strbuf errmsg = STRBUF_INIT;
208 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
209 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
210 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
211 errmsg.buf);
212 strbuf_release(&errmsg);
213 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
214 return 0;
215 }
216
217 if (!prefixcmp(var, "submodule."))
218 return parse_submodule_config_option(var, value);
219
220 return git_default_config(var, value, cb);
221}
222
223static char *quote_two(const char *one, const char *two)
224{
225 int need_one = quote_c_style(one, NULL, NULL, 1);
226 int need_two = quote_c_style(two, NULL, NULL, 1);
227 struct strbuf res = STRBUF_INIT;
228
229 if (need_one + need_two) {
230 strbuf_addch(&res, '"');
231 quote_c_style(one, &res, NULL, 1);
232 quote_c_style(two, &res, NULL, 1);
233 strbuf_addch(&res, '"');
234 } else {
235 strbuf_addstr(&res, one);
236 strbuf_addstr(&res, two);
237 }
238 return strbuf_detach(&res, NULL);
239}
240
241static const char *external_diff(void)
242{
243 static const char *external_diff_cmd = NULL;
244 static int done_preparing = 0;
245
246 if (done_preparing)
247 return external_diff_cmd;
248 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
249 if (!external_diff_cmd)
250 external_diff_cmd = external_diff_cmd_cfg;
251 done_preparing = 1;
252 return external_diff_cmd;
253}
254
255static struct diff_tempfile {
256 const char *name; /* filename external diff should read from */
257 char hex[41];
258 char mode[10];
259 char tmp_path[PATH_MAX];
260} diff_temp[2];
261
262typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
263
264struct emit_callback {
265 int color_diff;
266 unsigned ws_rule;
267 int blank_at_eof_in_preimage;
268 int blank_at_eof_in_postimage;
269 int lno_in_preimage;
270 int lno_in_postimage;
271 sane_truncate_fn truncate;
272 const char **label_path;
273 struct diff_words_data *diff_words;
274 struct diff_options *opt;
275 int *found_changesp;
276 struct strbuf *header;
277};
278
279static int count_lines(const char *data, int size)
280{
281 int count, ch, completely_empty = 1, nl_just_seen = 0;
282 count = 0;
283 while (0 < size--) {
284 ch = *data++;
285 if (ch == '\n') {
286 count++;
287 nl_just_seen = 1;
288 completely_empty = 0;
289 }
290 else {
291 nl_just_seen = 0;
292 completely_empty = 0;
293 }
294 }
295 if (completely_empty)
296 return 0;
297 if (!nl_just_seen)
298 count++; /* no trailing newline */
299 return count;
300}
301
302static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
303{
304 if (!DIFF_FILE_VALID(one)) {
305 mf->ptr = (char *)""; /* does not matter */
306 mf->size = 0;
307 return 0;
308 }
309 else if (diff_populate_filespec(one, 0))
310 return -1;
311
312 mf->ptr = one->data;
313 mf->size = one->size;
314 return 0;
315}
316
317/* like fill_mmfile, but only for size, so we can avoid retrieving blob */
318static unsigned long diff_filespec_size(struct diff_filespec *one)
319{
320 if (!DIFF_FILE_VALID(one))
321 return 0;
322 diff_populate_filespec(one, 1);
323 return one->size;
324}
325
326static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
327{
328 char *ptr = mf->ptr;
329 long size = mf->size;
330 int cnt = 0;
331
332 if (!size)
333 return cnt;
334 ptr += size - 1; /* pointing at the very end */
335 if (*ptr != '\n')
336 ; /* incomplete line */
337 else
338 ptr--; /* skip the last LF */
339 while (mf->ptr < ptr) {
340 char *prev_eol;
341 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
342 if (*prev_eol == '\n')
343 break;
344 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
345 break;
346 cnt++;
347 ptr = prev_eol - 1;
348 }
349 return cnt;
350}
351
352static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
353 struct emit_callback *ecbdata)
354{
355 int l1, l2, at;
356 unsigned ws_rule = ecbdata->ws_rule;
357 l1 = count_trailing_blank(mf1, ws_rule);
358 l2 = count_trailing_blank(mf2, ws_rule);
359 if (l2 <= l1) {
360 ecbdata->blank_at_eof_in_preimage = 0;
361 ecbdata->blank_at_eof_in_postimage = 0;
362 return;
363 }
364 at = count_lines(mf1->ptr, mf1->size);
365 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
366
367 at = count_lines(mf2->ptr, mf2->size);
368 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
369}
370
371static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
372 int first, const char *line, int len)
373{
374 int has_trailing_newline, has_trailing_carriage_return;
375 int nofirst;
376 FILE *file = o->file;
377
378 if (o->output_prefix) {
379 struct strbuf *msg = NULL;
380 msg = o->output_prefix(o, o->output_prefix_data);
381 assert(msg);
382 fwrite(msg->buf, msg->len, 1, file);
383 }
384
385 if (len == 0) {
386 has_trailing_newline = (first == '\n');
387 has_trailing_carriage_return = (!has_trailing_newline &&
388 (first == '\r'));
389 nofirst = has_trailing_newline || has_trailing_carriage_return;
390 } else {
391 has_trailing_newline = (len > 0 && line[len-1] == '\n');
392 if (has_trailing_newline)
393 len--;
394 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
395 if (has_trailing_carriage_return)
396 len--;
397 nofirst = 0;
398 }
399
400 if (len || !nofirst) {
401 fputs(set, file);
402 if (!nofirst)
403 fputc(first, file);
404 fwrite(line, len, 1, file);
405 fputs(reset, file);
406 }
407 if (has_trailing_carriage_return)
408 fputc('\r', file);
409 if (has_trailing_newline)
410 fputc('\n', file);
411}
412
413static void emit_line(struct diff_options *o, const char *set, const char *reset,
414 const char *line, int len)
415{
416 emit_line_0(o, set, reset, line[0], line+1, len-1);
417}
418
419static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
420{
421 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
422 ecbdata->blank_at_eof_in_preimage &&
423 ecbdata->blank_at_eof_in_postimage &&
424 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
425 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
426 return 0;
427 return ws_blank_line(line, len, ecbdata->ws_rule);
428}
429
430static void emit_add_line(const char *reset,
431 struct emit_callback *ecbdata,
432 const char *line, int len)
433{
434 const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
435 const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
436
437 if (!*ws)
438 emit_line_0(ecbdata->opt, set, reset, '+', line, len);
439 else if (new_blank_line_at_eof(ecbdata, line, len))
440 /* Blank line at EOF - paint '+' as well */
441 emit_line_0(ecbdata->opt, ws, reset, '+', line, len);
442 else {
443 /* Emit just the prefix, then the rest. */
444 emit_line_0(ecbdata->opt, set, reset, '+', "", 0);
445 ws_check_emit(line, len, ecbdata->ws_rule,
446 ecbdata->opt->file, set, reset, ws);
447 }
448}
449
450static void emit_hunk_header(struct emit_callback *ecbdata,
451 const char *line, int len)
452{
453 const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
454 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
455 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
456 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
457 static const char atat[2] = { '@', '@' };
458 const char *cp, *ep;
459 struct strbuf msgbuf = STRBUF_INIT;
460 int org_len = len;
461 int i = 1;
462
463 /*
464 * As a hunk header must begin with "@@ -<old>, +<new> @@",
465 * it always is at least 10 bytes long.
466 */
467 if (len < 10 ||
468 memcmp(line, atat, 2) ||
469 !(ep = memmem(line + 2, len - 2, atat, 2))) {
470 emit_line(ecbdata->opt, plain, reset, line, len);
471 return;
472 }
473 ep += 2; /* skip over @@ */
474
475 /* The hunk header in fraginfo color */
476 strbuf_add(&msgbuf, frag, strlen(frag));
477 strbuf_add(&msgbuf, line, ep - line);
478 strbuf_add(&msgbuf, reset, strlen(reset));
479
480 /*
481 * trailing "\r\n"
482 */
483 for ( ; i < 3; i++)
484 if (line[len - i] == '\r' || line[len - i] == '\n')
485 len--;
486
487 /* blank before the func header */
488 for (cp = ep; ep - line < len; ep++)
489 if (*ep != ' ' && *ep != '\t')
490 break;
491 if (ep != cp) {
492 strbuf_add(&msgbuf, plain, strlen(plain));
493 strbuf_add(&msgbuf, cp, ep - cp);
494 strbuf_add(&msgbuf, reset, strlen(reset));
495 }
496
497 if (ep < line + len) {
498 strbuf_add(&msgbuf, func, strlen(func));
499 strbuf_add(&msgbuf, ep, line + len - ep);
500 strbuf_add(&msgbuf, reset, strlen(reset));
501 }
502
503 strbuf_add(&msgbuf, line + len, org_len - len);
504 emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
505 strbuf_release(&msgbuf);
506}
507
508static struct diff_tempfile *claim_diff_tempfile(void) {
509 int i;
510 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
511 if (!diff_temp[i].name)
512 return diff_temp + i;
513 die("BUG: diff is failing to clean up its tempfiles");
514}
515
516static int remove_tempfile_installed;
517
518static void remove_tempfile(void)
519{
520 int i;
521 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
522 if (diff_temp[i].name == diff_temp[i].tmp_path)
523 unlink_or_warn(diff_temp[i].name);
524 diff_temp[i].name = NULL;
525 }
526}
527
528static void remove_tempfile_on_signal(int signo)
529{
530 remove_tempfile();
531 sigchain_pop(signo);
532 raise(signo);
533}
534
535static void print_line_count(FILE *file, int count)
536{
537 switch (count) {
538 case 0:
539 fprintf(file, "0,0");
540 break;
541 case 1:
542 fprintf(file, "1");
543 break;
544 default:
545 fprintf(file, "1,%d", count);
546 break;
547 }
548}
549
550static void emit_rewrite_lines(struct emit_callback *ecb,
551 int prefix, const char *data, int size)
552{
553 const char *endp = NULL;
554 static const char *nneof = " No newline at end of file\n";
555 const char *old = diff_get_color(ecb->color_diff, DIFF_FILE_OLD);
556 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
557
558 while (0 < size) {
559 int len;
560
561 endp = memchr(data, '\n', size);
562 len = endp ? (endp - data + 1) : size;
563 if (prefix != '+') {
564 ecb->lno_in_preimage++;
565 emit_line_0(ecb->opt, old, reset, '-',
566 data, len);
567 } else {
568 ecb->lno_in_postimage++;
569 emit_add_line(reset, ecb, data, len);
570 }
571 size -= len;
572 data += len;
573 }
574 if (!endp) {
575 const char *plain = diff_get_color(ecb->color_diff,
576 DIFF_PLAIN);
577 emit_line_0(ecb->opt, plain, reset, '\\',
578 nneof, strlen(nneof));
579 }
580}
581
582static void emit_rewrite_diff(const char *name_a,
583 const char *name_b,
584 struct diff_filespec *one,
585 struct diff_filespec *two,
586 struct userdiff_driver *textconv_one,
587 struct userdiff_driver *textconv_two,
588 struct diff_options *o)
589{
590 int lc_a, lc_b;
591 const char *name_a_tab, *name_b_tab;
592 const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
593 const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
594 const char *reset = diff_get_color(o->use_color, DIFF_RESET);
595 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
596 const char *a_prefix, *b_prefix;
597 char *data_one, *data_two;
598 size_t size_one, size_two;
599 struct emit_callback ecbdata;
600 char *line_prefix = "";
601 struct strbuf *msgbuf;
602
603 if (o && o->output_prefix) {
604 msgbuf = o->output_prefix(o, o->output_prefix_data);
605 line_prefix = msgbuf->buf;
606 }
607
608 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
609 a_prefix = o->b_prefix;
610 b_prefix = o->a_prefix;
611 } else {
612 a_prefix = o->a_prefix;
613 b_prefix = o->b_prefix;
614 }
615
616 name_a += (*name_a == '/');
617 name_b += (*name_b == '/');
618 name_a_tab = strchr(name_a, ' ') ? "\t" : "";
619 name_b_tab = strchr(name_b, ' ') ? "\t" : "";
620
621 strbuf_reset(&a_name);
622 strbuf_reset(&b_name);
623 quote_two_c_style(&a_name, a_prefix, name_a, 0);
624 quote_two_c_style(&b_name, b_prefix, name_b, 0);
625
626 size_one = fill_textconv(textconv_one, one, &data_one);
627 size_two = fill_textconv(textconv_two, two, &data_two);
628
629 memset(&ecbdata, 0, sizeof(ecbdata));
630 ecbdata.color_diff = want_color(o->use_color);
631 ecbdata.found_changesp = &o->found_changes;
632 ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
633 ecbdata.opt = o;
634 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
635 mmfile_t mf1, mf2;
636 mf1.ptr = (char *)data_one;
637 mf2.ptr = (char *)data_two;
638 mf1.size = size_one;
639 mf2.size = size_two;
640 check_blank_at_eof(&mf1, &mf2, &ecbdata);
641 }
642 ecbdata.lno_in_preimage = 1;
643 ecbdata.lno_in_postimage = 1;
644
645 lc_a = count_lines(data_one, size_one);
646 lc_b = count_lines(data_two, size_two);
647 fprintf(o->file,
648 "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
649 line_prefix, metainfo, a_name.buf, name_a_tab, reset,
650 line_prefix, metainfo, b_name.buf, name_b_tab, reset,
651 line_prefix, fraginfo);
652 if (!o->irreversible_delete)
653 print_line_count(o->file, lc_a);
654 else
655 fprintf(o->file, "?,?");
656 fprintf(o->file, " +");
657 print_line_count(o->file, lc_b);
658 fprintf(o->file, " @@%s\n", reset);
659 if (lc_a && !o->irreversible_delete)
660 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
661 if (lc_b)
662 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
663 if (textconv_one)
664 free((char *)data_one);
665 if (textconv_two)
666 free((char *)data_two);
667}
668
669struct diff_words_buffer {
670 mmfile_t text;
671 long alloc;
672 struct diff_words_orig {
673 const char *begin, *end;
674 } *orig;
675 int orig_nr, orig_alloc;
676};
677
678static void diff_words_append(char *line, unsigned long len,
679 struct diff_words_buffer *buffer)
680{
681 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
682 line++;
683 len--;
684 memcpy(buffer->text.ptr + buffer->text.size, line, len);
685 buffer->text.size += len;
686 buffer->text.ptr[buffer->text.size] = '\0';
687}
688
689struct diff_words_style_elem {
690 const char *prefix;
691 const char *suffix;
692 const char *color; /* NULL; filled in by the setup code if
693 * color is enabled */
694};
695
696struct diff_words_style {
697 enum diff_words_type type;
698 struct diff_words_style_elem new, old, ctx;
699 const char *newline;
700};
701
702static struct diff_words_style diff_words_styles[] = {
703 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
704 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
705 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
706};
707
708struct diff_words_data {
709 struct diff_words_buffer minus, plus;
710 const char *current_plus;
711 int last_minus;
712 struct diff_options *opt;
713 regex_t *word_regex;
714 enum diff_words_type type;
715 struct diff_words_style *style;
716};
717
718static int fn_out_diff_words_write_helper(FILE *fp,
719 struct diff_words_style_elem *st_el,
720 const char *newline,
721 size_t count, const char *buf,
722 const char *line_prefix)
723{
724 int print = 0;
725
726 while (count) {
727 char *p = memchr(buf, '\n', count);
728 if (print)
729 fputs(line_prefix, fp);
730 if (p != buf) {
731 if (st_el->color && fputs(st_el->color, fp) < 0)
732 return -1;
733 if (fputs(st_el->prefix, fp) < 0 ||
734 fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
735 fputs(st_el->suffix, fp) < 0)
736 return -1;
737 if (st_el->color && *st_el->color
738 && fputs(GIT_COLOR_RESET, fp) < 0)
739 return -1;
740 }
741 if (!p)
742 return 0;
743 if (fputs(newline, fp) < 0)
744 return -1;
745 count -= p + 1 - buf;
746 buf = p + 1;
747 print = 1;
748 }
749 return 0;
750}
751
752/*
753 * '--color-words' algorithm can be described as:
754 *
755 * 1. collect a the minus/plus lines of a diff hunk, divided into
756 * minus-lines and plus-lines;
757 *
758 * 2. break both minus-lines and plus-lines into words and
759 * place them into two mmfile_t with one word for each line;
760 *
761 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
762 *
763 * And for the common parts of the both file, we output the plus side text.
764 * diff_words->current_plus is used to trace the current position of the plus file
765 * which printed. diff_words->last_minus is used to trace the last minus word
766 * printed.
767 *
768 * For '--graph' to work with '--color-words', we need to output the graph prefix
769 * on each line of color words output. Generally, there are two conditions on
770 * which we should output the prefix.
771 *
772 * 1. diff_words->last_minus == 0 &&
773 * diff_words->current_plus == diff_words->plus.text.ptr
774 *
775 * that is: the plus text must start as a new line, and if there is no minus
776 * word printed, a graph prefix must be printed.
777 *
778 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
779 * *(diff_words->current_plus - 1) == '\n'
780 *
781 * that is: a graph prefix must be printed following a '\n'
782 */
783static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
784{
785 if ((diff_words->last_minus == 0 &&
786 diff_words->current_plus == diff_words->plus.text.ptr) ||
787 (diff_words->current_plus > diff_words->plus.text.ptr &&
788 *(diff_words->current_plus - 1) == '\n')) {
789 return 1;
790 } else {
791 return 0;
792 }
793}
794
795static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
796{
797 struct diff_words_data *diff_words = priv;
798 struct diff_words_style *style = diff_words->style;
799 int minus_first, minus_len, plus_first, plus_len;
800 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
801 struct diff_options *opt = diff_words->opt;
802 struct strbuf *msgbuf;
803 char *line_prefix = "";
804
805 if (line[0] != '@' || parse_hunk_header(line, len,
806 &minus_first, &minus_len, &plus_first, &plus_len))
807 return;
808
809 assert(opt);
810 if (opt->output_prefix) {
811 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
812 line_prefix = msgbuf->buf;
813 }
814
815 /* POSIX requires that first be decremented by one if len == 0... */
816 if (minus_len) {
817 minus_begin = diff_words->minus.orig[minus_first].begin;
818 minus_end =
819 diff_words->minus.orig[minus_first + minus_len - 1].end;
820 } else
821 minus_begin = minus_end =
822 diff_words->minus.orig[minus_first].end;
823
824 if (plus_len) {
825 plus_begin = diff_words->plus.orig[plus_first].begin;
826 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
827 } else
828 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
829
830 if (color_words_output_graph_prefix(diff_words)) {
831 fputs(line_prefix, diff_words->opt->file);
832 }
833 if (diff_words->current_plus != plus_begin) {
834 fn_out_diff_words_write_helper(diff_words->opt->file,
835 &style->ctx, style->newline,
836 plus_begin - diff_words->current_plus,
837 diff_words->current_plus, line_prefix);
838 if (*(plus_begin - 1) == '\n')
839 fputs(line_prefix, diff_words->opt->file);
840 }
841 if (minus_begin != minus_end) {
842 fn_out_diff_words_write_helper(diff_words->opt->file,
843 &style->old, style->newline,
844 minus_end - minus_begin, minus_begin,
845 line_prefix);
846 }
847 if (plus_begin != plus_end) {
848 fn_out_diff_words_write_helper(diff_words->opt->file,
849 &style->new, style->newline,
850 plus_end - plus_begin, plus_begin,
851 line_prefix);
852 }
853
854 diff_words->current_plus = plus_end;
855 diff_words->last_minus = minus_first;
856}
857
858/* This function starts looking at *begin, and returns 0 iff a word was found. */
859static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
860 int *begin, int *end)
861{
862 if (word_regex && *begin < buffer->size) {
863 regmatch_t match[1];
864 if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
865 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
866 '\n', match[0].rm_eo - match[0].rm_so);
867 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
868 *begin += match[0].rm_so;
869 return *begin >= *end;
870 }
871 return -1;
872 }
873
874 /* find the next word */
875 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
876 (*begin)++;
877 if (*begin >= buffer->size)
878 return -1;
879
880 /* find the end of the word */
881 *end = *begin + 1;
882 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
883 (*end)++;
884
885 return 0;
886}
887
888/*
889 * This function splits the words in buffer->text, stores the list with
890 * newline separator into out, and saves the offsets of the original words
891 * in buffer->orig.
892 */
893static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
894 regex_t *word_regex)
895{
896 int i, j;
897 long alloc = 0;
898
899 out->size = 0;
900 out->ptr = NULL;
901
902 /* fake an empty "0th" word */
903 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
904 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
905 buffer->orig_nr = 1;
906
907 for (i = 0; i < buffer->text.size; i++) {
908 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
909 return;
910
911 /* store original boundaries */
912 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
913 buffer->orig_alloc);
914 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
915 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
916 buffer->orig_nr++;
917
918 /* store one word */
919 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
920 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
921 out->ptr[out->size + j - i] = '\n';
922 out->size += j - i + 1;
923
924 i = j - 1;
925 }
926}
927
928/* this executes the word diff on the accumulated buffers */
929static void diff_words_show(struct diff_words_data *diff_words)
930{
931 xpparam_t xpp;
932 xdemitconf_t xecfg;
933 mmfile_t minus, plus;
934 struct diff_words_style *style = diff_words->style;
935
936 struct diff_options *opt = diff_words->opt;
937 struct strbuf *msgbuf;
938 char *line_prefix = "";
939
940 assert(opt);
941 if (opt->output_prefix) {
942 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
943 line_prefix = msgbuf->buf;
944 }
945
946 /* special case: only removal */
947 if (!diff_words->plus.text.size) {
948 fputs(line_prefix, diff_words->opt->file);
949 fn_out_diff_words_write_helper(diff_words->opt->file,
950 &style->old, style->newline,
951 diff_words->minus.text.size,
952 diff_words->minus.text.ptr, line_prefix);
953 diff_words->minus.text.size = 0;
954 return;
955 }
956
957 diff_words->current_plus = diff_words->plus.text.ptr;
958 diff_words->last_minus = 0;
959
960 memset(&xpp, 0, sizeof(xpp));
961 memset(&xecfg, 0, sizeof(xecfg));
962 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
963 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
964 xpp.flags = 0;
965 /* as only the hunk header will be parsed, we need a 0-context */
966 xecfg.ctxlen = 0;
967 xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
968 &xpp, &xecfg);
969 free(minus.ptr);
970 free(plus.ptr);
971 if (diff_words->current_plus != diff_words->plus.text.ptr +
972 diff_words->plus.text.size) {
973 if (color_words_output_graph_prefix(diff_words))
974 fputs(line_prefix, diff_words->opt->file);
975 fn_out_diff_words_write_helper(diff_words->opt->file,
976 &style->ctx, style->newline,
977 diff_words->plus.text.ptr + diff_words->plus.text.size
978 - diff_words->current_plus, diff_words->current_plus,
979 line_prefix);
980 }
981 diff_words->minus.text.size = diff_words->plus.text.size = 0;
982}
983
984/* In "color-words" mode, show word-diff of words accumulated in the buffer */
985static void diff_words_flush(struct emit_callback *ecbdata)
986{
987 if (ecbdata->diff_words->minus.text.size ||
988 ecbdata->diff_words->plus.text.size)
989 diff_words_show(ecbdata->diff_words);
990}
991
992static void free_diff_words_data(struct emit_callback *ecbdata)
993{
994 if (ecbdata->diff_words) {
995 diff_words_flush(ecbdata);
996 free (ecbdata->diff_words->minus.text.ptr);
997 free (ecbdata->diff_words->minus.orig);
998 free (ecbdata->diff_words->plus.text.ptr);
999 free (ecbdata->diff_words->plus.orig);
1000 if (ecbdata->diff_words->word_regex) {
1001 regfree(ecbdata->diff_words->word_regex);
1002 free(ecbdata->diff_words->word_regex);
1003 }
1004 free(ecbdata->diff_words);
1005 ecbdata->diff_words = NULL;
1006 }
1007}
1008
1009const char *diff_get_color(int diff_use_color, enum color_diff ix)
1010{
1011 if (want_color(diff_use_color))
1012 return diff_colors[ix];
1013 return "";
1014}
1015
1016static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1017{
1018 const char *cp;
1019 unsigned long allot;
1020 size_t l = len;
1021
1022 if (ecb->truncate)
1023 return ecb->truncate(line, len);
1024 cp = line;
1025 allot = l;
1026 while (0 < l) {
1027 (void) utf8_width(&cp, &l);
1028 if (!cp)
1029 break; /* truncated in the middle? */
1030 }
1031 return allot - l;
1032}
1033
1034static void find_lno(const char *line, struct emit_callback *ecbdata)
1035{
1036 const char *p;
1037 ecbdata->lno_in_preimage = 0;
1038 ecbdata->lno_in_postimage = 0;
1039 p = strchr(line, '-');
1040 if (!p)
1041 return; /* cannot happen */
1042 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1043 p = strchr(p, '+');
1044 if (!p)
1045 return; /* cannot happen */
1046 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1047}
1048
1049static void fn_out_consume(void *priv, char *line, unsigned long len)
1050{
1051 struct emit_callback *ecbdata = priv;
1052 const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1053 const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
1054 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1055 struct diff_options *o = ecbdata->opt;
1056 char *line_prefix = "";
1057 struct strbuf *msgbuf;
1058
1059 if (o && o->output_prefix) {
1060 msgbuf = o->output_prefix(o, o->output_prefix_data);
1061 line_prefix = msgbuf->buf;
1062 }
1063
1064 if (ecbdata->header) {
1065 fprintf(ecbdata->opt->file, "%s", ecbdata->header->buf);
1066 strbuf_reset(ecbdata->header);
1067 ecbdata->header = NULL;
1068 }
1069 *(ecbdata->found_changesp) = 1;
1070
1071 if (ecbdata->label_path[0]) {
1072 const char *name_a_tab, *name_b_tab;
1073
1074 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1075 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1076
1077 fprintf(ecbdata->opt->file, "%s%s--- %s%s%s\n",
1078 line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1079 fprintf(ecbdata->opt->file, "%s%s+++ %s%s%s\n",
1080 line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1081 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1082 }
1083
1084 if (diff_suppress_blank_empty
1085 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1086 line[0] = '\n';
1087 len = 1;
1088 }
1089
1090 if (line[0] == '@') {
1091 if (ecbdata->diff_words)
1092 diff_words_flush(ecbdata);
1093 len = sane_truncate_line(ecbdata, line, len);
1094 find_lno(line, ecbdata);
1095 emit_hunk_header(ecbdata, line, len);
1096 if (line[len-1] != '\n')
1097 putc('\n', ecbdata->opt->file);
1098 return;
1099 }
1100
1101 if (len < 1) {
1102 emit_line(ecbdata->opt, reset, reset, line, len);
1103 if (ecbdata->diff_words
1104 && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1105 fputs("~\n", ecbdata->opt->file);
1106 return;
1107 }
1108
1109 if (ecbdata->diff_words) {
1110 if (line[0] == '-') {
1111 diff_words_append(line, len,
1112 &ecbdata->diff_words->minus);
1113 return;
1114 } else if (line[0] == '+') {
1115 diff_words_append(line, len,
1116 &ecbdata->diff_words->plus);
1117 return;
1118 } else if (!prefixcmp(line, "\\ ")) {
1119 /*
1120 * Eat the "no newline at eof" marker as if we
1121 * saw a "+" or "-" line with nothing on it,
1122 * and return without diff_words_flush() to
1123 * defer processing. If this is the end of
1124 * preimage, more "+" lines may come after it.
1125 */
1126 return;
1127 }
1128 diff_words_flush(ecbdata);
1129 if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1130 emit_line(ecbdata->opt, plain, reset, line, len);
1131 fputs("~\n", ecbdata->opt->file);
1132 } else {
1133 /*
1134 * Skip the prefix character, if any. With
1135 * diff_suppress_blank_empty, there may be
1136 * none.
1137 */
1138 if (line[0] != '\n') {
1139 line++;
1140 len--;
1141 }
1142 emit_line(ecbdata->opt, plain, reset, line, len);
1143 }
1144 return;
1145 }
1146
1147 if (line[0] != '+') {
1148 const char *color =
1149 diff_get_color(ecbdata->color_diff,
1150 line[0] == '-' ? DIFF_FILE_OLD : DIFF_PLAIN);
1151 ecbdata->lno_in_preimage++;
1152 if (line[0] == ' ')
1153 ecbdata->lno_in_postimage++;
1154 emit_line(ecbdata->opt, color, reset, line, len);
1155 } else {
1156 ecbdata->lno_in_postimage++;
1157 emit_add_line(reset, ecbdata, line + 1, len - 1);
1158 }
1159}
1160
1161static char *pprint_rename(const char *a, const char *b)
1162{
1163 const char *old = a;
1164 const char *new = b;
1165 struct strbuf name = STRBUF_INIT;
1166 int pfx_length, sfx_length;
1167 int len_a = strlen(a);
1168 int len_b = strlen(b);
1169 int a_midlen, b_midlen;
1170 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1171 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1172
1173 if (qlen_a || qlen_b) {
1174 quote_c_style(a, &name, NULL, 0);
1175 strbuf_addstr(&name, " => ");
1176 quote_c_style(b, &name, NULL, 0);
1177 return strbuf_detach(&name, NULL);
1178 }
1179
1180 /* Find common prefix */
1181 pfx_length = 0;
1182 while (*old && *new && *old == *new) {
1183 if (*old == '/')
1184 pfx_length = old - a + 1;
1185 old++;
1186 new++;
1187 }
1188
1189 /* Find common suffix */
1190 old = a + len_a;
1191 new = b + len_b;
1192 sfx_length = 0;
1193 while (a <= old && b <= new && *old == *new) {
1194 if (*old == '/')
1195 sfx_length = len_a - (old - a);
1196 old--;
1197 new--;
1198 }
1199
1200 /*
1201 * pfx{mid-a => mid-b}sfx
1202 * {pfx-a => pfx-b}sfx
1203 * pfx{sfx-a => sfx-b}
1204 * name-a => name-b
1205 */
1206 a_midlen = len_a - pfx_length - sfx_length;
1207 b_midlen = len_b - pfx_length - sfx_length;
1208 if (a_midlen < 0)
1209 a_midlen = 0;
1210 if (b_midlen < 0)
1211 b_midlen = 0;
1212
1213 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1214 if (pfx_length + sfx_length) {
1215 strbuf_add(&name, a, pfx_length);
1216 strbuf_addch(&name, '{');
1217 }
1218 strbuf_add(&name, a + pfx_length, a_midlen);
1219 strbuf_addstr(&name, " => ");
1220 strbuf_add(&name, b + pfx_length, b_midlen);
1221 if (pfx_length + sfx_length) {
1222 strbuf_addch(&name, '}');
1223 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1224 }
1225 return strbuf_detach(&name, NULL);
1226}
1227
1228struct diffstat_t {
1229 int nr;
1230 int alloc;
1231 struct diffstat_file {
1232 char *from_name;
1233 char *name;
1234 char *print_name;
1235 unsigned is_unmerged:1;
1236 unsigned is_binary:1;
1237 unsigned is_renamed:1;
1238 uintmax_t added, deleted;
1239 } **files;
1240};
1241
1242static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1243 const char *name_a,
1244 const char *name_b)
1245{
1246 struct diffstat_file *x;
1247 x = xcalloc(sizeof (*x), 1);
1248 if (diffstat->nr == diffstat->alloc) {
1249 diffstat->alloc = alloc_nr(diffstat->alloc);
1250 diffstat->files = xrealloc(diffstat->files,
1251 diffstat->alloc * sizeof(x));
1252 }
1253 diffstat->files[diffstat->nr++] = x;
1254 if (name_b) {
1255 x->from_name = xstrdup(name_a);
1256 x->name = xstrdup(name_b);
1257 x->is_renamed = 1;
1258 }
1259 else {
1260 x->from_name = NULL;
1261 x->name = xstrdup(name_a);
1262 }
1263 return x;
1264}
1265
1266static void diffstat_consume(void *priv, char *line, unsigned long len)
1267{
1268 struct diffstat_t *diffstat = priv;
1269 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1270
1271 if (line[0] == '+')
1272 x->added++;
1273 else if (line[0] == '-')
1274 x->deleted++;
1275}
1276
1277const char mime_boundary_leader[] = "------------";
1278
1279static int scale_linear(int it, int width, int max_change)
1280{
1281 if (!it)
1282 return 0;
1283 /*
1284 * make sure that at least one '-' or '+' is printed if
1285 * there is any change to this path. The easiest way is to
1286 * scale linearly as if the alloted width is one column shorter
1287 * than it is, and then add 1 to the result.
1288 */
1289 return 1 + (it * (width - 1) / max_change);
1290}
1291
1292static void show_name(FILE *file,
1293 const char *prefix, const char *name, int len)
1294{
1295 fprintf(file, " %s%-*s |", prefix, len, name);
1296}
1297
1298static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1299{
1300 if (cnt <= 0)
1301 return;
1302 fprintf(file, "%s", set);
1303 while (cnt--)
1304 putc(ch, file);
1305 fprintf(file, "%s", reset);
1306}
1307
1308static void fill_print_name(struct diffstat_file *file)
1309{
1310 char *pname;
1311
1312 if (file->print_name)
1313 return;
1314
1315 if (!file->is_renamed) {
1316 struct strbuf buf = STRBUF_INIT;
1317 if (quote_c_style(file->name, &buf, NULL, 0)) {
1318 pname = strbuf_detach(&buf, NULL);
1319 } else {
1320 pname = file->name;
1321 strbuf_release(&buf);
1322 }
1323 } else {
1324 pname = pprint_rename(file->from_name, file->name);
1325 }
1326 file->print_name = pname;
1327}
1328
1329int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1330{
1331 struct strbuf sb = STRBUF_INIT;
1332 int ret;
1333
1334 if (!files) {
1335 assert(insertions == 0 && deletions == 0);
1336 return fputs(_(" 0 files changed\n"), fp);
1337 }
1338
1339 strbuf_addf(&sb,
1340 Q_(" %d file changed", " %d files changed", files),
1341 files);
1342
1343 /*
1344 * For binary diff, the caller may want to print "x files
1345 * changed" with insertions == 0 && deletions == 0.
1346 *
1347 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1348 * is probably less confusing (i.e skip over "2 files changed
1349 * but nothing about added/removed lines? Is this a bug in Git?").
1350 */
1351 if (insertions || deletions == 0) {
1352 /*
1353 * TRANSLATORS: "+" in (+) is a line addition marker;
1354 * do not translate it.
1355 */
1356 strbuf_addf(&sb,
1357 Q_(", %d insertion(+)", ", %d insertions(+)",
1358 insertions),
1359 insertions);
1360 }
1361
1362 if (deletions || insertions == 0) {
1363 /*
1364 * TRANSLATORS: "-" in (-) is a line removal marker;
1365 * do not translate it.
1366 */
1367 strbuf_addf(&sb,
1368 Q_(", %d deletion(-)", ", %d deletions(-)",
1369 deletions),
1370 deletions);
1371 }
1372 strbuf_addch(&sb, '\n');
1373 ret = fputs(sb.buf, fp);
1374 strbuf_release(&sb);
1375 return ret;
1376}
1377
1378static void show_stats(struct diffstat_t *data, struct diff_options *options)
1379{
1380 int i, len, add, del, adds = 0, dels = 0;
1381 uintmax_t max_change = 0, max_len = 0;
1382 int total_files = data->nr;
1383 int width, name_width, graph_width, number_width = 4, count;
1384 const char *reset, *add_c, *del_c;
1385 const char *line_prefix = "";
1386 int extra_shown = 0;
1387 struct strbuf *msg = NULL;
1388
1389 if (data->nr == 0)
1390 return;
1391
1392 if (options->output_prefix) {
1393 msg = options->output_prefix(options, options->output_prefix_data);
1394 line_prefix = msg->buf;
1395 }
1396
1397 count = options->stat_count ? options->stat_count : data->nr;
1398
1399 reset = diff_get_color_opt(options, DIFF_RESET);
1400 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1401 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1402
1403 /*
1404 * Find the longest filename and max number of changes
1405 */
1406 for (i = 0; (i < count) && (i < data->nr); i++) {
1407 struct diffstat_file *file = data->files[i];
1408 uintmax_t change = file->added + file->deleted;
1409 if (!data->files[i]->is_renamed &&
1410 (change == 0)) {
1411 count++; /* not shown == room for one more */
1412 continue;
1413 }
1414 fill_print_name(file);
1415 len = strlen(file->print_name);
1416 if (max_len < len)
1417 max_len = len;
1418
1419 if (file->is_binary || file->is_unmerged)
1420 continue;
1421 if (max_change < change)
1422 max_change = change;
1423 }
1424 count = i; /* min(count, data->nr) */
1425
1426 /*
1427 * We have width = stat_width or term_columns() columns total.
1428 * We want a maximum of min(max_len, stat_name_width) for the name part.
1429 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1430 * We also need 1 for " " and 4 + decimal_width(max_change)
1431 * for " | NNNN " and one the empty column at the end, altogether
1432 * 6 + decimal_width(max_change).
1433 *
1434 * If there's not enough space, we will use the smaller of
1435 * stat_name_width (if set) and 5/8*width for the filename,
1436 * and the rest for constant elements + graph part, but no more
1437 * than stat_graph_width for the graph part.
1438 * (5/8 gives 50 for filename and 30 for the constant parts + graph
1439 * for the standard terminal size).
1440 *
1441 * In other words: stat_width limits the maximum width, and
1442 * stat_name_width fixes the maximum width of the filename,
1443 * and is also used to divide available columns if there
1444 * aren't enough.
1445 */
1446
1447 if (options->stat_width == -1)
1448 width = term_columns();
1449 else
1450 width = options->stat_width ? options->stat_width : 80;
1451
1452 if (options->stat_graph_width == -1)
1453 options->stat_graph_width = diff_stat_graph_width;
1454
1455 /*
1456 * Guarantee 3/8*16==6 for the graph part
1457 * and 5/8*16==10 for the filename part
1458 */
1459 if (width < 16 + 6 + number_width)
1460 width = 16 + 6 + number_width;
1461
1462 /*
1463 * First assign sizes that are wanted, ignoring available width.
1464 */
1465 graph_width = (options->stat_graph_width &&
1466 options->stat_graph_width < max_change) ?
1467 options->stat_graph_width : max_change;
1468 name_width = (options->stat_name_width > 0 &&
1469 options->stat_name_width < max_len) ?
1470 options->stat_name_width : max_len;
1471
1472 /*
1473 * Adjust adjustable widths not to exceed maximum width
1474 */
1475 if (name_width + number_width + 6 + graph_width > width) {
1476 if (graph_width > width * 3/8 - number_width - 6)
1477 graph_width = width * 3/8 - number_width - 6;
1478 if (options->stat_graph_width &&
1479 graph_width > options->stat_graph_width)
1480 graph_width = options->stat_graph_width;
1481 if (name_width > width - number_width - 6 - graph_width)
1482 name_width = width - number_width - 6 - graph_width;
1483 else
1484 graph_width = width - number_width - 6 - name_width;
1485 }
1486
1487 /*
1488 * From here name_width is the width of the name area,
1489 * and graph_width is the width of the graph area.
1490 * max_change is used to scale graph properly.
1491 */
1492 for (i = 0; i < count; i++) {
1493 const char *prefix = "";
1494 char *name = data->files[i]->print_name;
1495 uintmax_t added = data->files[i]->added;
1496 uintmax_t deleted = data->files[i]->deleted;
1497 int name_len;
1498
1499 if (!data->files[i]->is_renamed &&
1500 (added + deleted == 0)) {
1501 total_files--;
1502 continue;
1503 }
1504 /*
1505 * "scale" the filename
1506 */
1507 len = name_width;
1508 name_len = strlen(name);
1509 if (name_width < name_len) {
1510 char *slash;
1511 prefix = "...";
1512 len -= 3;
1513 name += name_len - len;
1514 slash = strchr(name, '/');
1515 if (slash)
1516 name = slash;
1517 }
1518
1519 if (data->files[i]->is_binary) {
1520 fprintf(options->file, "%s", line_prefix);
1521 show_name(options->file, prefix, name, len);
1522 fprintf(options->file, " Bin ");
1523 fprintf(options->file, "%s%"PRIuMAX"%s",
1524 del_c, deleted, reset);
1525 fprintf(options->file, " -> ");
1526 fprintf(options->file, "%s%"PRIuMAX"%s",
1527 add_c, added, reset);
1528 fprintf(options->file, " bytes");
1529 fprintf(options->file, "\n");
1530 continue;
1531 }
1532 else if (data->files[i]->is_unmerged) {
1533 fprintf(options->file, "%s", line_prefix);
1534 show_name(options->file, prefix, name, len);
1535 fprintf(options->file, " Unmerged\n");
1536 continue;
1537 }
1538
1539 /*
1540 * scale the add/delete
1541 */
1542 add = added;
1543 del = deleted;
1544 adds += add;
1545 dels += del;
1546
1547 if (graph_width <= max_change) {
1548 int total = add + del;
1549
1550 total = scale_linear(add + del, graph_width, max_change);
1551 if (total < 2 && add && del)
1552 /* width >= 2 due to the sanity check */
1553 total = 2;
1554 if (add < del) {
1555 add = scale_linear(add, graph_width, max_change);
1556 del = total - add;
1557 } else {
1558 del = scale_linear(del, graph_width, max_change);
1559 add = total - del;
1560 }
1561 }
1562 fprintf(options->file, "%s", line_prefix);
1563 show_name(options->file, prefix, name, len);
1564 fprintf(options->file, "%5"PRIuMAX"%s", added + deleted,
1565 added + deleted ? " " : "");
1566 show_graph(options->file, '+', add, add_c, reset);
1567 show_graph(options->file, '-', del, del_c, reset);
1568 fprintf(options->file, "\n");
1569 }
1570 for (i = count; i < data->nr; i++) {
1571 uintmax_t added = data->files[i]->added;
1572 uintmax_t deleted = data->files[i]->deleted;
1573 if (!data->files[i]->is_renamed &&
1574 (added + deleted == 0)) {
1575 total_files--;
1576 continue;
1577 }
1578 adds += added;
1579 dels += deleted;
1580 if (!extra_shown)
1581 fprintf(options->file, "%s ...\n", line_prefix);
1582 extra_shown = 1;
1583 }
1584 fprintf(options->file, "%s", line_prefix);
1585 print_stat_summary(options->file, total_files, adds, dels);
1586}
1587
1588static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1589{
1590 int i, adds = 0, dels = 0, total_files = data->nr;
1591
1592 if (data->nr == 0)
1593 return;
1594
1595 for (i = 0; i < data->nr; i++) {
1596 if (!data->files[i]->is_binary &&
1597 !data->files[i]->is_unmerged) {
1598 int added = data->files[i]->added;
1599 int deleted= data->files[i]->deleted;
1600 if (!data->files[i]->is_renamed &&
1601 (added + deleted == 0)) {
1602 total_files--;
1603 } else {
1604 adds += added;
1605 dels += deleted;
1606 }
1607 }
1608 }
1609 if (options->output_prefix) {
1610 struct strbuf *msg = NULL;
1611 msg = options->output_prefix(options,
1612 options->output_prefix_data);
1613 fprintf(options->file, "%s", msg->buf);
1614 }
1615 print_stat_summary(options->file, total_files, adds, dels);
1616}
1617
1618static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1619{
1620 int i;
1621
1622 if (data->nr == 0)
1623 return;
1624
1625 for (i = 0; i < data->nr; i++) {
1626 struct diffstat_file *file = data->files[i];
1627
1628 if (options->output_prefix) {
1629 struct strbuf *msg = NULL;
1630 msg = options->output_prefix(options,
1631 options->output_prefix_data);
1632 fprintf(options->file, "%s", msg->buf);
1633 }
1634
1635 if (file->is_binary)
1636 fprintf(options->file, "-\t-\t");
1637 else
1638 fprintf(options->file,
1639 "%"PRIuMAX"\t%"PRIuMAX"\t",
1640 file->added, file->deleted);
1641 if (options->line_termination) {
1642 fill_print_name(file);
1643 if (!file->is_renamed)
1644 write_name_quoted(file->name, options->file,
1645 options->line_termination);
1646 else {
1647 fputs(file->print_name, options->file);
1648 putc(options->line_termination, options->file);
1649 }
1650 } else {
1651 if (file->is_renamed) {
1652 putc('\0', options->file);
1653 write_name_quoted(file->from_name, options->file, '\0');
1654 }
1655 write_name_quoted(file->name, options->file, '\0');
1656 }
1657 }
1658}
1659
1660struct dirstat_file {
1661 const char *name;
1662 unsigned long changed;
1663};
1664
1665struct dirstat_dir {
1666 struct dirstat_file *files;
1667 int alloc, nr, permille, cumulative;
1668};
1669
1670static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1671 unsigned long changed, const char *base, int baselen)
1672{
1673 unsigned long this_dir = 0;
1674 unsigned int sources = 0;
1675 const char *line_prefix = "";
1676 struct strbuf *msg = NULL;
1677
1678 if (opt->output_prefix) {
1679 msg = opt->output_prefix(opt, opt->output_prefix_data);
1680 line_prefix = msg->buf;
1681 }
1682
1683 while (dir->nr) {
1684 struct dirstat_file *f = dir->files;
1685 int namelen = strlen(f->name);
1686 unsigned long this;
1687 char *slash;
1688
1689 if (namelen < baselen)
1690 break;
1691 if (memcmp(f->name, base, baselen))
1692 break;
1693 slash = strchr(f->name + baselen, '/');
1694 if (slash) {
1695 int newbaselen = slash + 1 - f->name;
1696 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1697 sources++;
1698 } else {
1699 this = f->changed;
1700 dir->files++;
1701 dir->nr--;
1702 sources += 2;
1703 }
1704 this_dir += this;
1705 }
1706
1707 /*
1708 * We don't report dirstat's for
1709 * - the top level
1710 * - or cases where everything came from a single directory
1711 * under this directory (sources == 1).
1712 */
1713 if (baselen && sources != 1) {
1714 if (this_dir) {
1715 int permille = this_dir * 1000 / changed;
1716 if (permille >= dir->permille) {
1717 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1718 permille / 10, permille % 10, baselen, base);
1719 if (!dir->cumulative)
1720 return 0;
1721 }
1722 }
1723 }
1724 return this_dir;
1725}
1726
1727static int dirstat_compare(const void *_a, const void *_b)
1728{
1729 const struct dirstat_file *a = _a;
1730 const struct dirstat_file *b = _b;
1731 return strcmp(a->name, b->name);
1732}
1733
1734static void show_dirstat(struct diff_options *options)
1735{
1736 int i;
1737 unsigned long changed;
1738 struct dirstat_dir dir;
1739 struct diff_queue_struct *q = &diff_queued_diff;
1740
1741 dir.files = NULL;
1742 dir.alloc = 0;
1743 dir.nr = 0;
1744 dir.permille = options->dirstat_permille;
1745 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1746
1747 changed = 0;
1748 for (i = 0; i < q->nr; i++) {
1749 struct diff_filepair *p = q->queue[i];
1750 const char *name;
1751 unsigned long copied, added, damage;
1752 int content_changed;
1753
1754 name = p->two->path ? p->two->path : p->one->path;
1755
1756 if (p->one->sha1_valid && p->two->sha1_valid)
1757 content_changed = hashcmp(p->one->sha1, p->two->sha1);
1758 else
1759 content_changed = 1;
1760
1761 if (!content_changed) {
1762 /*
1763 * The SHA1 has not changed, so pre-/post-content is
1764 * identical. We can therefore skip looking at the
1765 * file contents altogether.
1766 */
1767 damage = 0;
1768 goto found_damage;
1769 }
1770
1771 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1772 /*
1773 * In --dirstat-by-file mode, we don't really need to
1774 * look at the actual file contents at all.
1775 * The fact that the SHA1 changed is enough for us to
1776 * add this file to the list of results
1777 * (with each file contributing equal damage).
1778 */
1779 damage = 1;
1780 goto found_damage;
1781 }
1782
1783 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1784 diff_populate_filespec(p->one, 0);
1785 diff_populate_filespec(p->two, 0);
1786 diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1787 &copied, &added);
1788 diff_free_filespec_data(p->one);
1789 diff_free_filespec_data(p->two);
1790 } else if (DIFF_FILE_VALID(p->one)) {
1791 diff_populate_filespec(p->one, 1);
1792 copied = added = 0;
1793 diff_free_filespec_data(p->one);
1794 } else if (DIFF_FILE_VALID(p->two)) {
1795 diff_populate_filespec(p->two, 1);
1796 copied = 0;
1797 added = p->two->size;
1798 diff_free_filespec_data(p->two);
1799 } else
1800 continue;
1801
1802 /*
1803 * Original minus copied is the removed material,
1804 * added is the new material. They are both damages
1805 * made to the preimage.
1806 * If the resulting damage is zero, we know that
1807 * diffcore_count_changes() considers the two entries to
1808 * be identical, but since content_changed is true, we
1809 * know that there must have been _some_ kind of change,
1810 * so we force all entries to have damage > 0.
1811 */
1812 damage = (p->one->size - copied) + added;
1813 if (!damage)
1814 damage = 1;
1815
1816found_damage:
1817 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1818 dir.files[dir.nr].name = name;
1819 dir.files[dir.nr].changed = damage;
1820 changed += damage;
1821 dir.nr++;
1822 }
1823
1824 /* This can happen even with many files, if everything was renames */
1825 if (!changed)
1826 return;
1827
1828 /* Show all directories with more than x% of the changes */
1829 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1830 gather_dirstat(options, &dir, changed, "", 0);
1831}
1832
1833static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
1834{
1835 int i;
1836 unsigned long changed;
1837 struct dirstat_dir dir;
1838
1839 if (data->nr == 0)
1840 return;
1841
1842 dir.files = NULL;
1843 dir.alloc = 0;
1844 dir.nr = 0;
1845 dir.permille = options->dirstat_permille;
1846 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1847
1848 changed = 0;
1849 for (i = 0; i < data->nr; i++) {
1850 struct diffstat_file *file = data->files[i];
1851 unsigned long damage = file->added + file->deleted;
1852 if (file->is_binary)
1853 /*
1854 * binary files counts bytes, not lines. Must find some
1855 * way to normalize binary bytes vs. textual lines.
1856 * The following heuristic assumes that there are 64
1857 * bytes per "line".
1858 * This is stupid and ugly, but very cheap...
1859 */
1860 damage = (damage + 63) / 64;
1861 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1862 dir.files[dir.nr].name = file->name;
1863 dir.files[dir.nr].changed = damage;
1864 changed += damage;
1865 dir.nr++;
1866 }
1867
1868 /* This can happen even with many files, if everything was renames */
1869 if (!changed)
1870 return;
1871
1872 /* Show all directories with more than x% of the changes */
1873 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1874 gather_dirstat(options, &dir, changed, "", 0);
1875}
1876
1877static void free_diffstat_info(struct diffstat_t *diffstat)
1878{
1879 int i;
1880 for (i = 0; i < diffstat->nr; i++) {
1881 struct diffstat_file *f = diffstat->files[i];
1882 if (f->name != f->print_name)
1883 free(f->print_name);
1884 free(f->name);
1885 free(f->from_name);
1886 free(f);
1887 }
1888 free(diffstat->files);
1889}
1890
1891struct checkdiff_t {
1892 const char *filename;
1893 int lineno;
1894 int conflict_marker_size;
1895 struct diff_options *o;
1896 unsigned ws_rule;
1897 unsigned status;
1898};
1899
1900static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
1901{
1902 char firstchar;
1903 int cnt;
1904
1905 if (len < marker_size + 1)
1906 return 0;
1907 firstchar = line[0];
1908 switch (firstchar) {
1909 case '=': case '>': case '<': case '|':
1910 break;
1911 default:
1912 return 0;
1913 }
1914 for (cnt = 1; cnt < marker_size; cnt++)
1915 if (line[cnt] != firstchar)
1916 return 0;
1917 /* line[1] thru line[marker_size-1] are same as firstchar */
1918 if (len < marker_size + 1 || !isspace(line[marker_size]))
1919 return 0;
1920 return 1;
1921}
1922
1923static void checkdiff_consume(void *priv, char *line, unsigned long len)
1924{
1925 struct checkdiff_t *data = priv;
1926 int marker_size = data->conflict_marker_size;
1927 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
1928 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
1929 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
1930 char *err;
1931 char *line_prefix = "";
1932 struct strbuf *msgbuf;
1933
1934 assert(data->o);
1935 if (data->o->output_prefix) {
1936 msgbuf = data->o->output_prefix(data->o,
1937 data->o->output_prefix_data);
1938 line_prefix = msgbuf->buf;
1939 }
1940
1941 if (line[0] == '+') {
1942 unsigned bad;
1943 data->lineno++;
1944 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
1945 data->status |= 1;
1946 fprintf(data->o->file,
1947 "%s%s:%d: leftover conflict marker\n",
1948 line_prefix, data->filename, data->lineno);
1949 }
1950 bad = ws_check(line + 1, len - 1, data->ws_rule);
1951 if (!bad)
1952 return;
1953 data->status |= bad;
1954 err = whitespace_error_string(bad);
1955 fprintf(data->o->file, "%s%s:%d: %s.\n",
1956 line_prefix, data->filename, data->lineno, err);
1957 free(err);
1958 emit_line(data->o, set, reset, line, 1);
1959 ws_check_emit(line + 1, len - 1, data->ws_rule,
1960 data->o->file, set, reset, ws);
1961 } else if (line[0] == ' ') {
1962 data->lineno++;
1963 } else if (line[0] == '@') {
1964 char *plus = strchr(line, '+');
1965 if (plus)
1966 data->lineno = strtol(plus, NULL, 10) - 1;
1967 else
1968 die("invalid diff");
1969 }
1970}
1971
1972static unsigned char *deflate_it(char *data,
1973 unsigned long size,
1974 unsigned long *result_size)
1975{
1976 int bound;
1977 unsigned char *deflated;
1978 git_zstream stream;
1979
1980 memset(&stream, 0, sizeof(stream));
1981 git_deflate_init(&stream, zlib_compression_level);
1982 bound = git_deflate_bound(&stream, size);
1983 deflated = xmalloc(bound);
1984 stream.next_out = deflated;
1985 stream.avail_out = bound;
1986
1987 stream.next_in = (unsigned char *)data;
1988 stream.avail_in = size;
1989 while (git_deflate(&stream, Z_FINISH) == Z_OK)
1990 ; /* nothing */
1991 git_deflate_end(&stream);
1992 *result_size = stream.total_out;
1993 return deflated;
1994}
1995
1996static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
1997{
1998 void *cp;
1999 void *delta;
2000 void *deflated;
2001 void *data;
2002 unsigned long orig_size;
2003 unsigned long delta_size;
2004 unsigned long deflate_size;
2005 unsigned long data_size;
2006
2007 /* We could do deflated delta, or we could do just deflated two,
2008 * whichever is smaller.
2009 */
2010 delta = NULL;
2011 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2012 if (one->size && two->size) {
2013 delta = diff_delta(one->ptr, one->size,
2014 two->ptr, two->size,
2015 &delta_size, deflate_size);
2016 if (delta) {
2017 void *to_free = delta;
2018 orig_size = delta_size;
2019 delta = deflate_it(delta, delta_size, &delta_size);
2020 free(to_free);
2021 }
2022 }
2023
2024 if (delta && delta_size < deflate_size) {
2025 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2026 free(deflated);
2027 data = delta;
2028 data_size = delta_size;
2029 }
2030 else {
2031 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2032 free(delta);
2033 data = deflated;
2034 data_size = deflate_size;
2035 }
2036
2037 /* emit data encoded in base85 */
2038 cp = data;
2039 while (data_size) {
2040 int bytes = (52 < data_size) ? 52 : data_size;
2041 char line[70];
2042 data_size -= bytes;
2043 if (bytes <= 26)
2044 line[0] = bytes + 'A' - 1;
2045 else
2046 line[0] = bytes - 26 + 'a' - 1;
2047 encode_85(line + 1, cp, bytes);
2048 cp = (char *) cp + bytes;
2049 fprintf(file, "%s", prefix);
2050 fputs(line, file);
2051 fputc('\n', file);
2052 }
2053 fprintf(file, "%s\n", prefix);
2054 free(data);
2055}
2056
2057static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
2058{
2059 fprintf(file, "%sGIT binary patch\n", prefix);
2060 emit_binary_diff_body(file, one, two, prefix);
2061 emit_binary_diff_body(file, two, one, prefix);
2062}
2063
2064static void diff_filespec_load_driver(struct diff_filespec *one)
2065{
2066 /* Use already-loaded driver */
2067 if (one->driver)
2068 return;
2069
2070 if (S_ISREG(one->mode))
2071 one->driver = userdiff_find_by_path(one->path);
2072
2073 /* Fallback to default settings */
2074 if (!one->driver)
2075 one->driver = userdiff_find_by_name("default");
2076}
2077
2078int diff_filespec_is_binary(struct diff_filespec *one)
2079{
2080 if (one->is_binary == -1) {
2081 diff_filespec_load_driver(one);
2082 if (one->driver->binary != -1)
2083 one->is_binary = one->driver->binary;
2084 else {
2085 if (!one->data && DIFF_FILE_VALID(one))
2086 diff_populate_filespec(one, 0);
2087 if (one->data)
2088 one->is_binary = buffer_is_binary(one->data,
2089 one->size);
2090 if (one->is_binary == -1)
2091 one->is_binary = 0;
2092 }
2093 }
2094 return one->is_binary;
2095}
2096
2097static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2098{
2099 diff_filespec_load_driver(one);
2100 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2101}
2102
2103static const char *userdiff_word_regex(struct diff_filespec *one)
2104{
2105 diff_filespec_load_driver(one);
2106 return one->driver->word_regex;
2107}
2108
2109void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2110{
2111 if (!options->a_prefix)
2112 options->a_prefix = a;
2113 if (!options->b_prefix)
2114 options->b_prefix = b;
2115}
2116
2117struct userdiff_driver *get_textconv(struct diff_filespec *one)
2118{
2119 if (!DIFF_FILE_VALID(one))
2120 return NULL;
2121
2122 diff_filespec_load_driver(one);
2123 return userdiff_get_textconv(one->driver);
2124}
2125
2126static void builtin_diff(const char *name_a,
2127 const char *name_b,
2128 struct diff_filespec *one,
2129 struct diff_filespec *two,
2130 const char *xfrm_msg,
2131 int must_show_header,
2132 struct diff_options *o,
2133 int complete_rewrite)
2134{
2135 mmfile_t mf1, mf2;
2136 const char *lbl[2];
2137 char *a_one, *b_two;
2138 const char *set = diff_get_color_opt(o, DIFF_METAINFO);
2139 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2140 const char *a_prefix, *b_prefix;
2141 struct userdiff_driver *textconv_one = NULL;
2142 struct userdiff_driver *textconv_two = NULL;
2143 struct strbuf header = STRBUF_INIT;
2144 struct strbuf *msgbuf;
2145 char *line_prefix = "";
2146
2147 if (o->output_prefix) {
2148 msgbuf = o->output_prefix(o, o->output_prefix_data);
2149 line_prefix = msgbuf->buf;
2150 }
2151
2152 if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2153 (!one->mode || S_ISGITLINK(one->mode)) &&
2154 (!two->mode || S_ISGITLINK(two->mode))) {
2155 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2156 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2157 show_submodule_summary(o->file, one ? one->path : two->path,
2158 one->sha1, two->sha1, two->dirty_submodule,
2159 del, add, reset);
2160 return;
2161 }
2162
2163 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2164 textconv_one = get_textconv(one);
2165 textconv_two = get_textconv(two);
2166 }
2167
2168 diff_set_mnemonic_prefix(o, "a/", "b/");
2169 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2170 a_prefix = o->b_prefix;
2171 b_prefix = o->a_prefix;
2172 } else {
2173 a_prefix = o->a_prefix;
2174 b_prefix = o->b_prefix;
2175 }
2176
2177 /* Never use a non-valid filename anywhere if at all possible */
2178 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2179 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2180
2181 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2182 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2183 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2184 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2185 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, set, a_one, b_two, reset);
2186 if (lbl[0][0] == '/') {
2187 /* /dev/null */
2188 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, set, two->mode, reset);
2189 if (xfrm_msg)
2190 strbuf_addstr(&header, xfrm_msg);
2191 must_show_header = 1;
2192 }
2193 else if (lbl[1][0] == '/') {
2194 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, set, one->mode, reset);
2195 if (xfrm_msg)
2196 strbuf_addstr(&header, xfrm_msg);
2197 must_show_header = 1;
2198 }
2199 else {
2200 if (one->mode != two->mode) {
2201 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, set, one->mode, reset);
2202 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, set, two->mode, reset);
2203 must_show_header = 1;
2204 }
2205 if (xfrm_msg)
2206 strbuf_addstr(&header, xfrm_msg);
2207
2208 /*
2209 * we do not run diff between different kind
2210 * of objects.
2211 */
2212 if ((one->mode ^ two->mode) & S_IFMT)
2213 goto free_ab_and_return;
2214 if (complete_rewrite &&
2215 (textconv_one || !diff_filespec_is_binary(one)) &&
2216 (textconv_two || !diff_filespec_is_binary(two))) {
2217 fprintf(o->file, "%s", header.buf);
2218 strbuf_reset(&header);
2219 emit_rewrite_diff(name_a, name_b, one, two,
2220 textconv_one, textconv_two, o);
2221 o->found_changes = 1;
2222 goto free_ab_and_return;
2223 }
2224 }
2225
2226 if (o->irreversible_delete && lbl[1][0] == '/') {
2227 fprintf(o->file, "%s", header.buf);
2228 strbuf_reset(&header);
2229 goto free_ab_and_return;
2230 } else if (!DIFF_OPT_TST(o, TEXT) &&
2231 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2232 (!textconv_two && diff_filespec_is_binary(two)) )) {
2233 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2234 die("unable to read files to diff");
2235 /* Quite common confusing case */
2236 if (mf1.size == mf2.size &&
2237 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2238 if (must_show_header)
2239 fprintf(o->file, "%s", header.buf);
2240 goto free_ab_and_return;
2241 }
2242 fprintf(o->file, "%s", header.buf);
2243 strbuf_reset(&header);
2244 if (DIFF_OPT_TST(o, BINARY))
2245 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2246 else
2247 fprintf(o->file, "%sBinary files %s and %s differ\n",
2248 line_prefix, lbl[0], lbl[1]);
2249 o->found_changes = 1;
2250 } else {
2251 /* Crazy xdl interfaces.. */
2252 const char *diffopts = getenv("GIT_DIFF_OPTS");
2253 xpparam_t xpp;
2254 xdemitconf_t xecfg;
2255 struct emit_callback ecbdata;
2256 const struct userdiff_funcname *pe;
2257
2258 if (must_show_header) {
2259 fprintf(o->file, "%s", header.buf);
2260 strbuf_reset(&header);
2261 }
2262
2263 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2264 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2265
2266 pe = diff_funcname_pattern(one);
2267 if (!pe)
2268 pe = diff_funcname_pattern(two);
2269
2270 memset(&xpp, 0, sizeof(xpp));
2271 memset(&xecfg, 0, sizeof(xecfg));
2272 memset(&ecbdata, 0, sizeof(ecbdata));
2273 ecbdata.label_path = lbl;
2274 ecbdata.color_diff = want_color(o->use_color);
2275 ecbdata.found_changesp = &o->found_changes;
2276 ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
2277 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2278 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2279 ecbdata.opt = o;
2280 ecbdata.header = header.len ? &header : NULL;
2281 xpp.flags = o->xdl_opts;
2282 xecfg.ctxlen = o->context;
2283 xecfg.interhunkctxlen = o->interhunkcontext;
2284 xecfg.flags = XDL_EMIT_FUNCNAMES;
2285 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2286 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2287 if (pe)
2288 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2289 if (!diffopts)
2290 ;
2291 else if (!prefixcmp(diffopts, "--unified="))
2292 xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
2293 else if (!prefixcmp(diffopts, "-u"))
2294 xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
2295 if (o->word_diff) {
2296 int i;
2297
2298 ecbdata.diff_words =
2299 xcalloc(1, sizeof(struct diff_words_data));
2300 ecbdata.diff_words->type = o->word_diff;
2301 ecbdata.diff_words->opt = o;
2302 if (!o->word_regex)
2303 o->word_regex = userdiff_word_regex(one);
2304 if (!o->word_regex)
2305 o->word_regex = userdiff_word_regex(two);
2306 if (!o->word_regex)
2307 o->word_regex = diff_word_regex_cfg;
2308 if (o->word_regex) {
2309 ecbdata.diff_words->word_regex = (regex_t *)
2310 xmalloc(sizeof(regex_t));
2311 if (regcomp(ecbdata.diff_words->word_regex,
2312 o->word_regex,
2313 REG_EXTENDED | REG_NEWLINE))
2314 die ("Invalid regular expression: %s",
2315 o->word_regex);
2316 }
2317 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
2318 if (o->word_diff == diff_words_styles[i].type) {
2319 ecbdata.diff_words->style =
2320 &diff_words_styles[i];
2321 break;
2322 }
2323 }
2324 if (want_color(o->use_color)) {
2325 struct diff_words_style *st = ecbdata.diff_words->style;
2326 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
2327 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
2328 st->ctx.color = diff_get_color_opt(o, DIFF_PLAIN);
2329 }
2330 }
2331 xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2332 &xpp, &xecfg);
2333 if (o->word_diff)
2334 free_diff_words_data(&ecbdata);
2335 if (textconv_one)
2336 free(mf1.ptr);
2337 if (textconv_two)
2338 free(mf2.ptr);
2339 xdiff_clear_find_func(&xecfg);
2340 }
2341
2342 free_ab_and_return:
2343 strbuf_release(&header);
2344 diff_free_filespec_data(one);
2345 diff_free_filespec_data(two);
2346 free(a_one);
2347 free(b_two);
2348 return;
2349}
2350
2351static void builtin_diffstat(const char *name_a, const char *name_b,
2352 struct diff_filespec *one,
2353 struct diff_filespec *two,
2354 struct diffstat_t *diffstat,
2355 struct diff_options *o,
2356 int complete_rewrite)
2357{
2358 mmfile_t mf1, mf2;
2359 struct diffstat_file *data;
2360
2361 data = diffstat_add(diffstat, name_a, name_b);
2362
2363 if (!one || !two) {
2364 data->is_unmerged = 1;
2365 return;
2366 }
2367
2368 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2369 data->is_binary = 1;
2370 data->added = diff_filespec_size(two);
2371 data->deleted = diff_filespec_size(one);
2372 }
2373
2374 else if (complete_rewrite) {
2375 diff_populate_filespec(one, 0);
2376 diff_populate_filespec(two, 0);
2377 data->deleted = count_lines(one->data, one->size);
2378 data->added = count_lines(two->data, two->size);
2379 }
2380
2381 else {
2382 /* Crazy xdl interfaces.. */
2383 xpparam_t xpp;
2384 xdemitconf_t xecfg;
2385
2386 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2387 die("unable to read files to diff");
2388
2389 memset(&xpp, 0, sizeof(xpp));
2390 memset(&xecfg, 0, sizeof(xecfg));
2391 xpp.flags = o->xdl_opts;
2392 xecfg.ctxlen = o->context;
2393 xecfg.interhunkctxlen = o->interhunkcontext;
2394 xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2395 &xpp, &xecfg);
2396 }
2397
2398 diff_free_filespec_data(one);
2399 diff_free_filespec_data(two);
2400}
2401
2402static void builtin_checkdiff(const char *name_a, const char *name_b,
2403 const char *attr_path,
2404 struct diff_filespec *one,
2405 struct diff_filespec *two,
2406 struct diff_options *o)
2407{
2408 mmfile_t mf1, mf2;
2409 struct checkdiff_t data;
2410
2411 if (!two)
2412 return;
2413
2414 memset(&data, 0, sizeof(data));
2415 data.filename = name_b ? name_b : name_a;
2416 data.lineno = 0;
2417 data.o = o;
2418 data.ws_rule = whitespace_rule(attr_path);
2419 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2420
2421 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2422 die("unable to read files to diff");
2423
2424 /*
2425 * All the other codepaths check both sides, but not checking
2426 * the "old" side here is deliberate. We are checking the newly
2427 * introduced changes, and as long as the "new" side is text, we
2428 * can and should check what it introduces.
2429 */
2430 if (diff_filespec_is_binary(two))
2431 goto free_and_return;
2432 else {
2433 /* Crazy xdl interfaces.. */
2434 xpparam_t xpp;
2435 xdemitconf_t xecfg;
2436
2437 memset(&xpp, 0, sizeof(xpp));
2438 memset(&xecfg, 0, sizeof(xecfg));
2439 xecfg.ctxlen = 1; /* at least one context line */
2440 xpp.flags = 0;
2441 xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2442 &xpp, &xecfg);
2443
2444 if (data.ws_rule & WS_BLANK_AT_EOF) {
2445 struct emit_callback ecbdata;
2446 int blank_at_eof;
2447
2448 ecbdata.ws_rule = data.ws_rule;
2449 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2450 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2451
2452 if (blank_at_eof) {
2453 static char *err;
2454 if (!err)
2455 err = whitespace_error_string(WS_BLANK_AT_EOF);
2456 fprintf(o->file, "%s:%d: %s.\n",
2457 data.filename, blank_at_eof, err);
2458 data.status = 1; /* report errors */
2459 }
2460 }
2461 }
2462 free_and_return:
2463 diff_free_filespec_data(one);
2464 diff_free_filespec_data(two);
2465 if (data.status)
2466 DIFF_OPT_SET(o, CHECK_FAILED);
2467}
2468
2469struct diff_filespec *alloc_filespec(const char *path)
2470{
2471 int namelen = strlen(path);
2472 struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
2473
2474 memset(spec, 0, sizeof(*spec));
2475 spec->path = (char *)(spec + 1);
2476 memcpy(spec->path, path, namelen+1);
2477 spec->count = 1;
2478 spec->is_binary = -1;
2479 return spec;
2480}
2481
2482void free_filespec(struct diff_filespec *spec)
2483{
2484 if (!--spec->count) {
2485 diff_free_filespec_data(spec);
2486 free(spec);
2487 }
2488}
2489
2490void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2491 unsigned short mode)
2492{
2493 if (mode) {
2494 spec->mode = canon_mode(mode);
2495 hashcpy(spec->sha1, sha1);
2496 spec->sha1_valid = !is_null_sha1(sha1);
2497 }
2498}
2499
2500/*
2501 * Given a name and sha1 pair, if the index tells us the file in
2502 * the work tree has that object contents, return true, so that
2503 * prepare_temp_file() does not have to inflate and extract.
2504 */
2505static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2506{
2507 struct cache_entry *ce;
2508 struct stat st;
2509 int pos, len;
2510
2511 /*
2512 * We do not read the cache ourselves here, because the
2513 * benchmark with my previous version that always reads cache
2514 * shows that it makes things worse for diff-tree comparing
2515 * two linux-2.6 kernel trees in an already checked out work
2516 * tree. This is because most diff-tree comparisons deal with
2517 * only a small number of files, while reading the cache is
2518 * expensive for a large project, and its cost outweighs the
2519 * savings we get by not inflating the object to a temporary
2520 * file. Practically, this code only helps when we are used
2521 * by diff-cache --cached, which does read the cache before
2522 * calling us.
2523 */
2524 if (!active_cache)
2525 return 0;
2526
2527 /* We want to avoid the working directory if our caller
2528 * doesn't need the data in a normal file, this system
2529 * is rather slow with its stat/open/mmap/close syscalls,
2530 * and the object is contained in a pack file. The pack
2531 * is probably already open and will be faster to obtain
2532 * the data through than the working directory. Loose
2533 * objects however would tend to be slower as they need
2534 * to be individually opened and inflated.
2535 */
2536 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2537 return 0;
2538
2539 len = strlen(name);
2540 pos = cache_name_pos(name, len);
2541 if (pos < 0)
2542 return 0;
2543 ce = active_cache[pos];
2544
2545 /*
2546 * This is not the sha1 we are looking for, or
2547 * unreusable because it is not a regular file.
2548 */
2549 if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2550 return 0;
2551
2552 /*
2553 * If ce is marked as "assume unchanged", there is no
2554 * guarantee that work tree matches what we are looking for.
2555 */
2556 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2557 return 0;
2558
2559 /*
2560 * If ce matches the file in the work tree, we can reuse it.
2561 */
2562 if (ce_uptodate(ce) ||
2563 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2564 return 1;
2565
2566 return 0;
2567}
2568
2569static int populate_from_stdin(struct diff_filespec *s)
2570{
2571 struct strbuf buf = STRBUF_INIT;
2572 size_t size = 0;
2573
2574 if (strbuf_read(&buf, 0, 0) < 0)
2575 return error("error while reading from stdin %s",
2576 strerror(errno));
2577
2578 s->should_munmap = 0;
2579 s->data = strbuf_detach(&buf, &size);
2580 s->size = size;
2581 s->should_free = 1;
2582 return 0;
2583}
2584
2585static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2586{
2587 int len;
2588 char *data = xmalloc(100), *dirty = "";
2589
2590 /* Are we looking at the work tree? */
2591 if (s->dirty_submodule)
2592 dirty = "-dirty";
2593
2594 len = snprintf(data, 100,
2595 "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2596 s->data = data;
2597 s->size = len;
2598 s->should_free = 1;
2599 if (size_only) {
2600 s->data = NULL;
2601 free(data);
2602 }
2603 return 0;
2604}
2605
2606/*
2607 * While doing rename detection and pickaxe operation, we may need to
2608 * grab the data for the blob (or file) for our own in-core comparison.
2609 * diff_filespec has data and size fields for this purpose.
2610 */
2611int diff_populate_filespec(struct diff_filespec *s, int size_only)
2612{
2613 int err = 0;
2614 if (!DIFF_FILE_VALID(s))
2615 die("internal error: asking to populate invalid file.");
2616 if (S_ISDIR(s->mode))
2617 return -1;
2618
2619 if (s->data)
2620 return 0;
2621
2622 if (size_only && 0 < s->size)
2623 return 0;
2624
2625 if (S_ISGITLINK(s->mode))
2626 return diff_populate_gitlink(s, size_only);
2627
2628 if (!s->sha1_valid ||
2629 reuse_worktree_file(s->path, s->sha1, 0)) {
2630 struct strbuf buf = STRBUF_INIT;
2631 struct stat st;
2632 int fd;
2633
2634 if (!strcmp(s->path, "-"))
2635 return populate_from_stdin(s);
2636
2637 if (lstat(s->path, &st) < 0) {
2638 if (errno == ENOENT) {
2639 err_empty:
2640 err = -1;
2641 empty:
2642 s->data = (char *)"";
2643 s->size = 0;
2644 return err;
2645 }
2646 }
2647 s->size = xsize_t(st.st_size);
2648 if (!s->size)
2649 goto empty;
2650 if (S_ISLNK(st.st_mode)) {
2651 struct strbuf sb = STRBUF_INIT;
2652
2653 if (strbuf_readlink(&sb, s->path, s->size))
2654 goto err_empty;
2655 s->size = sb.len;
2656 s->data = strbuf_detach(&sb, NULL);
2657 s->should_free = 1;
2658 return 0;
2659 }
2660 if (size_only)
2661 return 0;
2662 fd = open(s->path, O_RDONLY);
2663 if (fd < 0)
2664 goto err_empty;
2665 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2666 close(fd);
2667 s->should_munmap = 1;
2668
2669 /*
2670 * Convert from working tree format to canonical git format
2671 */
2672 if (convert_to_git(s->path, s->data, s->size, &buf, safe_crlf)) {
2673 size_t size = 0;
2674 munmap(s->data, s->size);
2675 s->should_munmap = 0;
2676 s->data = strbuf_detach(&buf, &size);
2677 s->size = size;
2678 s->should_free = 1;
2679 }
2680 }
2681 else {
2682 enum object_type type;
2683 if (size_only) {
2684 type = sha1_object_info(s->sha1, &s->size);
2685 if (type < 0)
2686 die("unable to read %s", sha1_to_hex(s->sha1));
2687 } else {
2688 s->data = read_sha1_file(s->sha1, &type, &s->size);
2689 if (!s->data)
2690 die("unable to read %s", sha1_to_hex(s->sha1));
2691 s->should_free = 1;
2692 }
2693 }
2694 return 0;
2695}
2696
2697void diff_free_filespec_blob(struct diff_filespec *s)
2698{
2699 if (s->should_free)
2700 free(s->data);
2701 else if (s->should_munmap)
2702 munmap(s->data, s->size);
2703
2704 if (s->should_free || s->should_munmap) {
2705 s->should_free = s->should_munmap = 0;
2706 s->data = NULL;
2707 }
2708}
2709
2710void diff_free_filespec_data(struct diff_filespec *s)
2711{
2712 diff_free_filespec_blob(s);
2713 free(s->cnt_data);
2714 s->cnt_data = NULL;
2715}
2716
2717static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2718 void *blob,
2719 unsigned long size,
2720 const unsigned char *sha1,
2721 int mode)
2722{
2723 int fd;
2724 struct strbuf buf = STRBUF_INIT;
2725 struct strbuf template = STRBUF_INIT;
2726 char *path_dup = xstrdup(path);
2727 const char *base = basename(path_dup);
2728
2729 /* Generate "XXXXXX_basename.ext" */
2730 strbuf_addstr(&template, "XXXXXX_");
2731 strbuf_addstr(&template, base);
2732
2733 fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
2734 strlen(base) + 1);
2735 if (fd < 0)
2736 die_errno("unable to create temp-file");
2737 if (convert_to_working_tree(path,
2738 (const char *)blob, (size_t)size, &buf)) {
2739 blob = buf.buf;
2740 size = buf.len;
2741 }
2742 if (write_in_full(fd, blob, size) != size)
2743 die_errno("unable to write temp-file");
2744 close(fd);
2745 temp->name = temp->tmp_path;
2746 strcpy(temp->hex, sha1_to_hex(sha1));
2747 temp->hex[40] = 0;
2748 sprintf(temp->mode, "%06o", mode);
2749 strbuf_release(&buf);
2750 strbuf_release(&template);
2751 free(path_dup);
2752}
2753
2754static struct diff_tempfile *prepare_temp_file(const char *name,
2755 struct diff_filespec *one)
2756{
2757 struct diff_tempfile *temp = claim_diff_tempfile();
2758
2759 if (!DIFF_FILE_VALID(one)) {
2760 not_a_valid_file:
2761 /* A '-' entry produces this for file-2, and
2762 * a '+' entry produces this for file-1.
2763 */
2764 temp->name = "/dev/null";
2765 strcpy(temp->hex, ".");
2766 strcpy(temp->mode, ".");
2767 return temp;
2768 }
2769
2770 if (!remove_tempfile_installed) {
2771 atexit(remove_tempfile);
2772 sigchain_push_common(remove_tempfile_on_signal);
2773 remove_tempfile_installed = 1;
2774 }
2775
2776 if (!one->sha1_valid ||
2777 reuse_worktree_file(name, one->sha1, 1)) {
2778 struct stat st;
2779 if (lstat(name, &st) < 0) {
2780 if (errno == ENOENT)
2781 goto not_a_valid_file;
2782 die_errno("stat(%s)", name);
2783 }
2784 if (S_ISLNK(st.st_mode)) {
2785 struct strbuf sb = STRBUF_INIT;
2786 if (strbuf_readlink(&sb, name, st.st_size) < 0)
2787 die_errno("readlink(%s)", name);
2788 prep_temp_blob(name, temp, sb.buf, sb.len,
2789 (one->sha1_valid ?
2790 one->sha1 : null_sha1),
2791 (one->sha1_valid ?
2792 one->mode : S_IFLNK));
2793 strbuf_release(&sb);
2794 }
2795 else {
2796 /* we can borrow from the file in the work tree */
2797 temp->name = name;
2798 if (!one->sha1_valid)
2799 strcpy(temp->hex, sha1_to_hex(null_sha1));
2800 else
2801 strcpy(temp->hex, sha1_to_hex(one->sha1));
2802 /* Even though we may sometimes borrow the
2803 * contents from the work tree, we always want
2804 * one->mode. mode is trustworthy even when
2805 * !(one->sha1_valid), as long as
2806 * DIFF_FILE_VALID(one).
2807 */
2808 sprintf(temp->mode, "%06o", one->mode);
2809 }
2810 return temp;
2811 }
2812 else {
2813 if (diff_populate_filespec(one, 0))
2814 die("cannot read data blob for %s", one->path);
2815 prep_temp_blob(name, temp, one->data, one->size,
2816 one->sha1, one->mode);
2817 }
2818 return temp;
2819}
2820
2821/* An external diff command takes:
2822 *
2823 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2824 * infile2 infile2-sha1 infile2-mode [ rename-to ]
2825 *
2826 */
2827static void run_external_diff(const char *pgm,
2828 const char *name,
2829 const char *other,
2830 struct diff_filespec *one,
2831 struct diff_filespec *two,
2832 const char *xfrm_msg,
2833 int complete_rewrite)
2834{
2835 const char *spawn_arg[10];
2836 int retval;
2837 const char **arg = &spawn_arg[0];
2838
2839 if (one && two) {
2840 struct diff_tempfile *temp_one, *temp_two;
2841 const char *othername = (other ? other : name);
2842 temp_one = prepare_temp_file(name, one);
2843 temp_two = prepare_temp_file(othername, two);
2844 *arg++ = pgm;
2845 *arg++ = name;
2846 *arg++ = temp_one->name;
2847 *arg++ = temp_one->hex;
2848 *arg++ = temp_one->mode;
2849 *arg++ = temp_two->name;
2850 *arg++ = temp_two->hex;
2851 *arg++ = temp_two->mode;
2852 if (other) {
2853 *arg++ = other;
2854 *arg++ = xfrm_msg;
2855 }
2856 } else {
2857 *arg++ = pgm;
2858 *arg++ = name;
2859 }
2860 *arg = NULL;
2861 fflush(NULL);
2862 retval = run_command_v_opt(spawn_arg, RUN_USING_SHELL);
2863 remove_tempfile();
2864 if (retval) {
2865 fprintf(stderr, "external diff died, stopping at %s.\n", name);
2866 exit(1);
2867 }
2868}
2869
2870static int similarity_index(struct diff_filepair *p)
2871{
2872 return p->score * 100 / MAX_SCORE;
2873}
2874
2875static void fill_metainfo(struct strbuf *msg,
2876 const char *name,
2877 const char *other,
2878 struct diff_filespec *one,
2879 struct diff_filespec *two,
2880 struct diff_options *o,
2881 struct diff_filepair *p,
2882 int *must_show_header,
2883 int use_color)
2884{
2885 const char *set = diff_get_color(use_color, DIFF_METAINFO);
2886 const char *reset = diff_get_color(use_color, DIFF_RESET);
2887 struct strbuf *msgbuf;
2888 char *line_prefix = "";
2889
2890 *must_show_header = 1;
2891 if (o->output_prefix) {
2892 msgbuf = o->output_prefix(o, o->output_prefix_data);
2893 line_prefix = msgbuf->buf;
2894 }
2895 strbuf_init(msg, PATH_MAX * 2 + 300);
2896 switch (p->status) {
2897 case DIFF_STATUS_COPIED:
2898 strbuf_addf(msg, "%s%ssimilarity index %d%%",
2899 line_prefix, set, similarity_index(p));
2900 strbuf_addf(msg, "%s\n%s%scopy from ",
2901 reset, line_prefix, set);
2902 quote_c_style(name, msg, NULL, 0);
2903 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
2904 quote_c_style(other, msg, NULL, 0);
2905 strbuf_addf(msg, "%s\n", reset);
2906 break;
2907 case DIFF_STATUS_RENAMED:
2908 strbuf_addf(msg, "%s%ssimilarity index %d%%",
2909 line_prefix, set, similarity_index(p));
2910 strbuf_addf(msg, "%s\n%s%srename from ",
2911 reset, line_prefix, set);
2912 quote_c_style(name, msg, NULL, 0);
2913 strbuf_addf(msg, "%s\n%s%srename to ",
2914 reset, line_prefix, set);
2915 quote_c_style(other, msg, NULL, 0);
2916 strbuf_addf(msg, "%s\n", reset);
2917 break;
2918 case DIFF_STATUS_MODIFIED:
2919 if (p->score) {
2920 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
2921 line_prefix,
2922 set, similarity_index(p), reset);
2923 break;
2924 }
2925 /* fallthru */
2926 default:
2927 *must_show_header = 0;
2928 }
2929 if (one && two && hashcmp(one->sha1, two->sha1)) {
2930 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2931
2932 if (DIFF_OPT_TST(o, BINARY)) {
2933 mmfile_t mf;
2934 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
2935 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
2936 abbrev = 40;
2937 }
2938 strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
2939 find_unique_abbrev(one->sha1, abbrev));
2940 strbuf_addstr(msg, find_unique_abbrev(two->sha1, abbrev));
2941 if (one->mode == two->mode)
2942 strbuf_addf(msg, " %06o", one->mode);
2943 strbuf_addf(msg, "%s\n", reset);
2944 }
2945}
2946
2947static void run_diff_cmd(const char *pgm,
2948 const char *name,
2949 const char *other,
2950 const char *attr_path,
2951 struct diff_filespec *one,
2952 struct diff_filespec *two,
2953 struct strbuf *msg,
2954 struct diff_options *o,
2955 struct diff_filepair *p)
2956{
2957 const char *xfrm_msg = NULL;
2958 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
2959 int must_show_header = 0;
2960
2961 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
2962 pgm = NULL;
2963 else {
2964 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
2965 if (drv && drv->external)
2966 pgm = drv->external;
2967 }
2968
2969 if (msg) {
2970 /*
2971 * don't use colors when the header is intended for an
2972 * external diff driver
2973 */
2974 fill_metainfo(msg, name, other, one, two, o, p,
2975 &must_show_header,
2976 want_color(o->use_color) && !pgm);
2977 xfrm_msg = msg->len ? msg->buf : NULL;
2978 }
2979
2980 if (pgm) {
2981 run_external_diff(pgm, name, other, one, two, xfrm_msg,
2982 complete_rewrite);
2983 return;
2984 }
2985 if (one && two)
2986 builtin_diff(name, other ? other : name,
2987 one, two, xfrm_msg, must_show_header,
2988 o, complete_rewrite);
2989 else
2990 fprintf(o->file, "* Unmerged path %s\n", name);
2991}
2992
2993static void diff_fill_sha1_info(struct diff_filespec *one)
2994{
2995 if (DIFF_FILE_VALID(one)) {
2996 if (!one->sha1_valid) {
2997 struct stat st;
2998 if (!strcmp(one->path, "-")) {
2999 hashcpy(one->sha1, null_sha1);
3000 return;
3001 }
3002 if (lstat(one->path, &st) < 0)
3003 die_errno("stat '%s'", one->path);
3004 if (index_path(one->sha1, one->path, &st, 0))
3005 die("cannot hash %s", one->path);
3006 }
3007 }
3008 else
3009 hashclr(one->sha1);
3010}
3011
3012static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3013{
3014 /* Strip the prefix but do not molest /dev/null and absolute paths */
3015 if (*namep && **namep != '/') {
3016 *namep += prefix_length;
3017 if (**namep == '/')
3018 ++*namep;
3019 }
3020 if (*otherp && **otherp != '/') {
3021 *otherp += prefix_length;
3022 if (**otherp == '/')
3023 ++*otherp;
3024 }
3025}
3026
3027static void run_diff(struct diff_filepair *p, struct diff_options *o)
3028{
3029 const char *pgm = external_diff();
3030 struct strbuf msg;
3031 struct diff_filespec *one = p->one;
3032 struct diff_filespec *two = p->two;
3033 const char *name;
3034 const char *other;
3035 const char *attr_path;
3036
3037 name = p->one->path;
3038 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3039 attr_path = name;
3040 if (o->prefix_length)
3041 strip_prefix(o->prefix_length, &name, &other);
3042
3043 if (DIFF_PAIR_UNMERGED(p)) {
3044 run_diff_cmd(pgm, name, NULL, attr_path,
3045 NULL, NULL, NULL, o, p);
3046 return;
3047 }
3048
3049 diff_fill_sha1_info(one);
3050 diff_fill_sha1_info(two);
3051
3052 if (!pgm &&
3053 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3054 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3055 /*
3056 * a filepair that changes between file and symlink
3057 * needs to be split into deletion and creation.
3058 */
3059 struct diff_filespec *null = alloc_filespec(two->path);
3060 run_diff_cmd(NULL, name, other, attr_path,
3061 one, null, &msg, o, p);
3062 free(null);
3063 strbuf_release(&msg);
3064
3065 null = alloc_filespec(one->path);
3066 run_diff_cmd(NULL, name, other, attr_path,
3067 null, two, &msg, o, p);
3068 free(null);
3069 }
3070 else
3071 run_diff_cmd(pgm, name, other, attr_path,
3072 one, two, &msg, o, p);
3073
3074 strbuf_release(&msg);
3075}
3076
3077static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3078 struct diffstat_t *diffstat)
3079{
3080 const char *name;
3081 const char *other;
3082 int complete_rewrite = 0;
3083
3084 if (DIFF_PAIR_UNMERGED(p)) {
3085 /* unmerged */
3086 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
3087 return;
3088 }
3089
3090 name = p->one->path;
3091 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3092
3093 if (o->prefix_length)
3094 strip_prefix(o->prefix_length, &name, &other);
3095
3096 diff_fill_sha1_info(p->one);
3097 diff_fill_sha1_info(p->two);
3098
3099 if (p->status == DIFF_STATUS_MODIFIED && p->score)
3100 complete_rewrite = 1;
3101 builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
3102}
3103
3104static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3105{
3106 const char *name;
3107 const char *other;
3108 const char *attr_path;
3109
3110 if (DIFF_PAIR_UNMERGED(p)) {
3111 /* unmerged */
3112 return;
3113 }
3114
3115 name = p->one->path;
3116 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3117 attr_path = other ? other : name;
3118
3119 if (o->prefix_length)
3120 strip_prefix(o->prefix_length, &name, &other);
3121
3122 diff_fill_sha1_info(p->one);
3123 diff_fill_sha1_info(p->two);
3124
3125 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3126}
3127
3128void diff_setup(struct diff_options *options)
3129{
3130 memcpy(options, &default_diff_options, sizeof(*options));
3131
3132 options->file = stdout;
3133
3134 options->line_termination = '\n';
3135 options->break_opt = -1;
3136 options->rename_limit = -1;
3137 options->dirstat_permille = diff_dirstat_permille_default;
3138 options->context = 3;
3139
3140 options->change = diff_change;
3141 options->add_remove = diff_addremove;
3142 options->use_color = diff_use_color_default;
3143 options->detect_rename = diff_detect_rename_default;
3144
3145 if (diff_no_prefix) {
3146 options->a_prefix = options->b_prefix = "";
3147 } else if (!diff_mnemonic_prefix) {
3148 options->a_prefix = "a/";
3149 options->b_prefix = "b/";
3150 }
3151}
3152
3153int diff_setup_done(struct diff_options *options)
3154{
3155 int count = 0;
3156
3157 if (options->output_format & DIFF_FORMAT_NAME)
3158 count++;
3159 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3160 count++;
3161 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3162 count++;
3163 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3164 count++;
3165 if (count > 1)
3166 die("--name-only, --name-status, --check and -s are mutually exclusive");
3167
3168 /*
3169 * Most of the time we can say "there are changes"
3170 * only by checking if there are changed paths, but
3171 * --ignore-whitespace* options force us to look
3172 * inside contents.
3173 */
3174
3175 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3176 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3177 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3178 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3179 else
3180 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3181
3182 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3183 options->detect_rename = DIFF_DETECT_COPY;
3184
3185 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3186 options->prefix = NULL;
3187 if (options->prefix)
3188 options->prefix_length = strlen(options->prefix);
3189 else
3190 options->prefix_length = 0;
3191
3192 if (options->output_format & (DIFF_FORMAT_NAME |
3193 DIFF_FORMAT_NAME_STATUS |
3194 DIFF_FORMAT_CHECKDIFF |
3195 DIFF_FORMAT_NO_OUTPUT))
3196 options->output_format &= ~(DIFF_FORMAT_RAW |
3197 DIFF_FORMAT_NUMSTAT |
3198 DIFF_FORMAT_DIFFSTAT |
3199 DIFF_FORMAT_SHORTSTAT |
3200 DIFF_FORMAT_DIRSTAT |
3201 DIFF_FORMAT_SUMMARY |
3202 DIFF_FORMAT_PATCH);
3203
3204 /*
3205 * These cases always need recursive; we do not drop caller-supplied
3206 * recursive bits for other formats here.
3207 */
3208 if (options->output_format & (DIFF_FORMAT_PATCH |
3209 DIFF_FORMAT_NUMSTAT |
3210 DIFF_FORMAT_DIFFSTAT |
3211 DIFF_FORMAT_SHORTSTAT |
3212 DIFF_FORMAT_DIRSTAT |
3213 DIFF_FORMAT_SUMMARY |
3214 DIFF_FORMAT_CHECKDIFF))
3215 DIFF_OPT_SET(options, RECURSIVE);
3216 /*
3217 * Also pickaxe would not work very well if you do not say recursive
3218 */
3219 if (options->pickaxe)
3220 DIFF_OPT_SET(options, RECURSIVE);
3221 /*
3222 * When patches are generated, submodules diffed against the work tree
3223 * must be checked for dirtiness too so it can be shown in the output
3224 */
3225 if (options->output_format & DIFF_FORMAT_PATCH)
3226 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3227
3228 if (options->detect_rename && options->rename_limit < 0)
3229 options->rename_limit = diff_rename_limit_default;
3230 if (options->setup & DIFF_SETUP_USE_CACHE) {
3231 if (!active_cache)
3232 /* read-cache does not die even when it fails
3233 * so it is safe for us to do this here. Also
3234 * it does not smudge active_cache or active_nr
3235 * when it fails, so we do not have to worry about
3236 * cleaning it up ourselves either.
3237 */
3238 read_cache();
3239 }
3240 if (options->abbrev <= 0 || 40 < options->abbrev)
3241 options->abbrev = 40; /* full */
3242
3243 /*
3244 * It does not make sense to show the first hit we happened
3245 * to have found. It does not make sense not to return with
3246 * exit code in such a case either.
3247 */
3248 if (DIFF_OPT_TST(options, QUICK)) {
3249 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3250 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3251 }
3252
3253 return 0;
3254}
3255
3256static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3257{
3258 char c, *eq;
3259 int len;
3260
3261 if (*arg != '-')
3262 return 0;
3263 c = *++arg;
3264 if (!c)
3265 return 0;
3266 if (c == arg_short) {
3267 c = *++arg;
3268 if (!c)
3269 return 1;
3270 if (val && isdigit(c)) {
3271 char *end;
3272 int n = strtoul(arg, &end, 10);
3273 if (*end)
3274 return 0;
3275 *val = n;
3276 return 1;
3277 }
3278 return 0;
3279 }
3280 if (c != '-')
3281 return 0;
3282 arg++;
3283 eq = strchr(arg, '=');
3284 if (eq)
3285 len = eq - arg;
3286 else
3287 len = strlen(arg);
3288 if (!len || strncmp(arg, arg_long, len))
3289 return 0;
3290 if (eq) {
3291 int n;
3292 char *end;
3293 if (!isdigit(*++eq))
3294 return 0;
3295 n = strtoul(eq, &end, 10);
3296 if (*end)
3297 return 0;
3298 *val = n;
3299 }
3300 return 1;
3301}
3302
3303static int diff_scoreopt_parse(const char *opt);
3304
3305static inline int short_opt(char opt, const char **argv,
3306 const char **optarg)
3307{
3308 const char *arg = argv[0];
3309 if (arg[0] != '-' || arg[1] != opt)
3310 return 0;
3311 if (arg[2] != '\0') {
3312 *optarg = arg + 2;
3313 return 1;
3314 }
3315 if (!argv[1])
3316 die("Option '%c' requires a value", opt);
3317 *optarg = argv[1];
3318 return 2;
3319}
3320
3321int parse_long_opt(const char *opt, const char **argv,
3322 const char **optarg)
3323{
3324 const char *arg = argv[0];
3325 if (arg[0] != '-' || arg[1] != '-')
3326 return 0;
3327 arg += strlen("--");
3328 if (prefixcmp(arg, opt))
3329 return 0;
3330 arg += strlen(opt);
3331 if (*arg == '=') { /* sticked form: --option=value */
3332 *optarg = arg + 1;
3333 return 1;
3334 }
3335 if (*arg != '\0')
3336 return 0;
3337 /* separate form: --option value */
3338 if (!argv[1])
3339 die("Option '--%s' requires a value", opt);
3340 *optarg = argv[1];
3341 return 2;
3342}
3343
3344static int stat_opt(struct diff_options *options, const char **av)
3345{
3346 const char *arg = av[0];
3347 char *end;
3348 int width = options->stat_width;
3349 int name_width = options->stat_name_width;
3350 int graph_width = options->stat_graph_width;
3351 int count = options->stat_count;
3352 int argcount = 1;
3353
3354 arg += strlen("--stat");
3355 end = (char *)arg;
3356
3357 switch (*arg) {
3358 case '-':
3359 if (!prefixcmp(arg, "-width")) {
3360 arg += strlen("-width");
3361 if (*arg == '=')
3362 width = strtoul(arg + 1, &end, 10);
3363 else if (!*arg && !av[1])
3364 die("Option '--stat-width' requires a value");
3365 else if (!*arg) {
3366 width = strtoul(av[1], &end, 10);
3367 argcount = 2;
3368 }
3369 } else if (!prefixcmp(arg, "-name-width")) {
3370 arg += strlen("-name-width");
3371 if (*arg == '=')
3372 name_width = strtoul(arg + 1, &end, 10);
3373 else if (!*arg && !av[1])
3374 die("Option '--stat-name-width' requires a value");
3375 else if (!*arg) {
3376 name_width = strtoul(av[1], &end, 10);
3377 argcount = 2;
3378 }
3379 } else if (!prefixcmp(arg, "-graph-width")) {
3380 arg += strlen("-graph-width");
3381 if (*arg == '=')
3382 graph_width = strtoul(arg + 1, &end, 10);
3383 else if (!*arg && !av[1])
3384 die("Option '--stat-graph-width' requires a value");
3385 else if (!*arg) {
3386 graph_width = strtoul(av[1], &end, 10);
3387 argcount = 2;
3388 }
3389 } else if (!prefixcmp(arg, "-count")) {
3390 arg += strlen("-count");
3391 if (*arg == '=')
3392 count = strtoul(arg + 1, &end, 10);
3393 else if (!*arg && !av[1])
3394 die("Option '--stat-count' requires a value");
3395 else if (!*arg) {
3396 count = strtoul(av[1], &end, 10);
3397 argcount = 2;
3398 }
3399 }
3400 break;
3401 case '=':
3402 width = strtoul(arg+1, &end, 10);
3403 if (*end == ',')
3404 name_width = strtoul(end+1, &end, 10);
3405 if (*end == ',')
3406 count = strtoul(end+1, &end, 10);
3407 }
3408
3409 /* Important! This checks all the error cases! */
3410 if (*end)
3411 return 0;
3412 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3413 options->stat_name_width = name_width;
3414 options->stat_graph_width = graph_width;
3415 options->stat_width = width;
3416 options->stat_count = count;
3417 return argcount;
3418}
3419
3420static int parse_dirstat_opt(struct diff_options *options, const char *params)
3421{
3422 struct strbuf errmsg = STRBUF_INIT;
3423 if (parse_dirstat_params(options, params, &errmsg))
3424 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3425 errmsg.buf);
3426 strbuf_release(&errmsg);
3427 /*
3428 * The caller knows a dirstat-related option is given from the command
3429 * line; allow it to say "return this_function();"
3430 */
3431 options->output_format |= DIFF_FORMAT_DIRSTAT;
3432 return 1;
3433}
3434
3435int diff_opt_parse(struct diff_options *options, const char **av, int ac)
3436{
3437 const char *arg = av[0];
3438 const char *optarg;
3439 int argcount;
3440
3441 /* Output format options */
3442 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch"))
3443 options->output_format |= DIFF_FORMAT_PATCH;
3444 else if (opt_arg(arg, 'U', "unified", &options->context))
3445 options->output_format |= DIFF_FORMAT_PATCH;
3446 else if (!strcmp(arg, "--raw"))
3447 options->output_format |= DIFF_FORMAT_RAW;
3448 else if (!strcmp(arg, "--patch-with-raw"))
3449 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
3450 else if (!strcmp(arg, "--numstat"))
3451 options->output_format |= DIFF_FORMAT_NUMSTAT;
3452 else if (!strcmp(arg, "--shortstat"))
3453 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3454 else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3455 return parse_dirstat_opt(options, "");
3456 else if (!prefixcmp(arg, "-X"))
3457 return parse_dirstat_opt(options, arg + 2);
3458 else if (!prefixcmp(arg, "--dirstat="))
3459 return parse_dirstat_opt(options, arg + 10);
3460 else if (!strcmp(arg, "--cumulative"))
3461 return parse_dirstat_opt(options, "cumulative");
3462 else if (!strcmp(arg, "--dirstat-by-file"))
3463 return parse_dirstat_opt(options, "files");
3464 else if (!prefixcmp(arg, "--dirstat-by-file=")) {
3465 parse_dirstat_opt(options, "files");
3466 return parse_dirstat_opt(options, arg + 18);
3467 }
3468 else if (!strcmp(arg, "--check"))
3469 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3470 else if (!strcmp(arg, "--summary"))
3471 options->output_format |= DIFF_FORMAT_SUMMARY;
3472 else if (!strcmp(arg, "--patch-with-stat"))
3473 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
3474 else if (!strcmp(arg, "--name-only"))
3475 options->output_format |= DIFF_FORMAT_NAME;
3476 else if (!strcmp(arg, "--name-status"))
3477 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3478 else if (!strcmp(arg, "-s"))
3479 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3480 else if (!prefixcmp(arg, "--stat"))
3481 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3482 return stat_opt(options, av);
3483
3484 /* renames options */
3485 else if (!prefixcmp(arg, "-B") || !prefixcmp(arg, "--break-rewrites=") ||
3486 !strcmp(arg, "--break-rewrites")) {
3487 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3488 return error("invalid argument to -B: %s", arg+2);
3489 }
3490 else if (!prefixcmp(arg, "-M") || !prefixcmp(arg, "--find-renames=") ||
3491 !strcmp(arg, "--find-renames")) {
3492 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3493 return error("invalid argument to -M: %s", arg+2);
3494 options->detect_rename = DIFF_DETECT_RENAME;
3495 }
3496 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3497 options->irreversible_delete = 1;
3498 }
3499 else if (!prefixcmp(arg, "-C") || !prefixcmp(arg, "--find-copies=") ||
3500 !strcmp(arg, "--find-copies")) {
3501 if (options->detect_rename == DIFF_DETECT_COPY)
3502 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3503 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3504 return error("invalid argument to -C: %s", arg+2);
3505 options->detect_rename = DIFF_DETECT_COPY;
3506 }
3507 else if (!strcmp(arg, "--no-renames"))
3508 options->detect_rename = 0;
3509 else if (!strcmp(arg, "--relative"))
3510 DIFF_OPT_SET(options, RELATIVE_NAME);
3511 else if (!prefixcmp(arg, "--relative=")) {
3512 DIFF_OPT_SET(options, RELATIVE_NAME);
3513 options->prefix = arg + 11;
3514 }
3515
3516 /* xdiff options */
3517 else if (!strcmp(arg, "--minimal"))
3518 DIFF_XDL_SET(options, NEED_MINIMAL);
3519 else if (!strcmp(arg, "--no-minimal"))
3520 DIFF_XDL_CLR(options, NEED_MINIMAL);
3521 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3522 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3523 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3524 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3525 else if (!strcmp(arg, "--ignore-space-at-eol"))
3526 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3527 else if (!strcmp(arg, "--patience"))
3528 DIFF_XDL_SET(options, PATIENCE_DIFF);
3529 else if (!strcmp(arg, "--histogram"))
3530 DIFF_XDL_SET(options, HISTOGRAM_DIFF);
3531
3532 /* flags options */
3533 else if (!strcmp(arg, "--binary")) {
3534 options->output_format |= DIFF_FORMAT_PATCH;
3535 DIFF_OPT_SET(options, BINARY);
3536 }
3537 else if (!strcmp(arg, "--full-index"))
3538 DIFF_OPT_SET(options, FULL_INDEX);
3539 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3540 DIFF_OPT_SET(options, TEXT);
3541 else if (!strcmp(arg, "-R"))
3542 DIFF_OPT_SET(options, REVERSE_DIFF);
3543 else if (!strcmp(arg, "--find-copies-harder"))
3544 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3545 else if (!strcmp(arg, "--follow"))
3546 DIFF_OPT_SET(options, FOLLOW_RENAMES);
3547 else if (!strcmp(arg, "--color"))
3548 options->use_color = 1;
3549 else if (!prefixcmp(arg, "--color=")) {
3550 int value = git_config_colorbool(NULL, arg+8);
3551 if (value < 0)
3552 return error("option `color' expects \"always\", \"auto\", or \"never\"");
3553 options->use_color = value;
3554 }
3555 else if (!strcmp(arg, "--no-color"))
3556 options->use_color = 0;
3557 else if (!strcmp(arg, "--color-words")) {
3558 options->use_color = 1;
3559 options->word_diff = DIFF_WORDS_COLOR;
3560 }
3561 else if (!prefixcmp(arg, "--color-words=")) {
3562 options->use_color = 1;
3563 options->word_diff = DIFF_WORDS_COLOR;
3564 options->word_regex = arg + 14;
3565 }
3566 else if (!strcmp(arg, "--word-diff")) {
3567 if (options->word_diff == DIFF_WORDS_NONE)
3568 options->word_diff = DIFF_WORDS_PLAIN;
3569 }
3570 else if (!prefixcmp(arg, "--word-diff=")) {
3571 const char *type = arg + 12;
3572 if (!strcmp(type, "plain"))
3573 options->word_diff = DIFF_WORDS_PLAIN;
3574 else if (!strcmp(type, "color")) {
3575 options->use_color = 1;
3576 options->word_diff = DIFF_WORDS_COLOR;
3577 }
3578 else if (!strcmp(type, "porcelain"))
3579 options->word_diff = DIFF_WORDS_PORCELAIN;
3580 else if (!strcmp(type, "none"))
3581 options->word_diff = DIFF_WORDS_NONE;
3582 else
3583 die("bad --word-diff argument: %s", type);
3584 }
3585 else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3586 if (options->word_diff == DIFF_WORDS_NONE)
3587 options->word_diff = DIFF_WORDS_PLAIN;
3588 options->word_regex = optarg;
3589 return argcount;
3590 }
3591 else if (!strcmp(arg, "--exit-code"))
3592 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3593 else if (!strcmp(arg, "--quiet"))
3594 DIFF_OPT_SET(options, QUICK);
3595 else if (!strcmp(arg, "--ext-diff"))
3596 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3597 else if (!strcmp(arg, "--no-ext-diff"))
3598 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3599 else if (!strcmp(arg, "--textconv"))
3600 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3601 else if (!strcmp(arg, "--no-textconv"))
3602 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3603 else if (!strcmp(arg, "--ignore-submodules")) {
3604 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3605 handle_ignore_submodules_arg(options, "all");
3606 } else if (!prefixcmp(arg, "--ignore-submodules=")) {
3607 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3608 handle_ignore_submodules_arg(options, arg + 20);
3609 } else if (!strcmp(arg, "--submodule"))
3610 DIFF_OPT_SET(options, SUBMODULE_LOG);
3611 else if (!prefixcmp(arg, "--submodule=")) {
3612 if (!strcmp(arg + 12, "log"))
3613 DIFF_OPT_SET(options, SUBMODULE_LOG);
3614 }
3615
3616 /* misc options */
3617 else if (!strcmp(arg, "-z"))
3618 options->line_termination = 0;
3619 else if ((argcount = short_opt('l', av, &optarg))) {
3620 options->rename_limit = strtoul(optarg, NULL, 10);
3621 return argcount;
3622 }
3623 else if ((argcount = short_opt('S', av, &optarg))) {
3624 options->pickaxe = optarg;
3625 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3626 return argcount;
3627 } else if ((argcount = short_opt('G', av, &optarg))) {
3628 options->pickaxe = optarg;
3629 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3630 return argcount;
3631 }
3632 else if (!strcmp(arg, "--pickaxe-all"))
3633 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3634 else if (!strcmp(arg, "--pickaxe-regex"))
3635 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3636 else if ((argcount = short_opt('O', av, &optarg))) {
3637 options->orderfile = optarg;
3638 return argcount;
3639 }
3640 else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3641 options->filter = optarg;
3642 return argcount;
3643 }
3644 else if (!strcmp(arg, "--abbrev"))
3645 options->abbrev = DEFAULT_ABBREV;
3646 else if (!prefixcmp(arg, "--abbrev=")) {
3647 options->abbrev = strtoul(arg + 9, NULL, 10);
3648 if (options->abbrev < MINIMUM_ABBREV)
3649 options->abbrev = MINIMUM_ABBREV;
3650 else if (40 < options->abbrev)
3651 options->abbrev = 40;
3652 }
3653 else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3654 options->a_prefix = optarg;
3655 return argcount;
3656 }
3657 else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
3658 options->b_prefix = optarg;
3659 return argcount;
3660 }
3661 else if (!strcmp(arg, "--no-prefix"))
3662 options->a_prefix = options->b_prefix = "";
3663 else if (opt_arg(arg, '\0', "inter-hunk-context",
3664 &options->interhunkcontext))
3665 ;
3666 else if (!strcmp(arg, "-W"))
3667 DIFF_OPT_SET(options, FUNCCONTEXT);
3668 else if (!strcmp(arg, "--function-context"))
3669 DIFF_OPT_SET(options, FUNCCONTEXT);
3670 else if (!strcmp(arg, "--no-function-context"))
3671 DIFF_OPT_CLR(options, FUNCCONTEXT);
3672 else if ((argcount = parse_long_opt("output", av, &optarg))) {
3673 options->file = fopen(optarg, "w");
3674 if (!options->file)
3675 die_errno("Could not open '%s'", optarg);
3676 options->close_file = 1;
3677 return argcount;
3678 } else
3679 return 0;
3680 return 1;
3681}
3682
3683int parse_rename_score(const char **cp_p)
3684{
3685 unsigned long num, scale;
3686 int ch, dot;
3687 const char *cp = *cp_p;
3688
3689 num = 0;
3690 scale = 1;
3691 dot = 0;
3692 for (;;) {
3693 ch = *cp;
3694 if ( !dot && ch == '.' ) {
3695 scale = 1;
3696 dot = 1;
3697 } else if ( ch == '%' ) {
3698 scale = dot ? scale*100 : 100;
3699 cp++; /* % is always at the end */
3700 break;
3701 } else if ( ch >= '0' && ch <= '9' ) {
3702 if ( scale < 100000 ) {
3703 scale *= 10;
3704 num = (num*10) + (ch-'0');
3705 }
3706 } else {
3707 break;
3708 }
3709 cp++;
3710 }
3711 *cp_p = cp;
3712
3713 /* user says num divided by scale and we say internally that
3714 * is MAX_SCORE * num / scale.
3715 */
3716 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
3717}
3718
3719static int diff_scoreopt_parse(const char *opt)
3720{
3721 int opt1, opt2, cmd;
3722
3723 if (*opt++ != '-')
3724 return -1;
3725 cmd = *opt++;
3726 if (cmd == '-') {
3727 /* convert the long-form arguments into short-form versions */
3728 if (!prefixcmp(opt, "break-rewrites")) {
3729 opt += strlen("break-rewrites");
3730 if (*opt == 0 || *opt++ == '=')
3731 cmd = 'B';
3732 } else if (!prefixcmp(opt, "find-copies")) {
3733 opt += strlen("find-copies");
3734 if (*opt == 0 || *opt++ == '=')
3735 cmd = 'C';
3736 } else if (!prefixcmp(opt, "find-renames")) {
3737 opt += strlen("find-renames");
3738 if (*opt == 0 || *opt++ == '=')
3739 cmd = 'M';
3740 }
3741 }
3742 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
3743 return -1; /* that is not a -M, -C nor -B option */
3744
3745 opt1 = parse_rename_score(&opt);
3746 if (cmd != 'B')
3747 opt2 = 0;
3748 else {
3749 if (*opt == 0)
3750 opt2 = 0;
3751 else if (*opt != '/')
3752 return -1; /* we expect -B80/99 or -B80 */
3753 else {
3754 opt++;
3755 opt2 = parse_rename_score(&opt);
3756 }
3757 }
3758 if (*opt != 0)
3759 return -1;
3760 return opt1 | (opt2 << 16);
3761}
3762
3763struct diff_queue_struct diff_queued_diff;
3764
3765void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
3766{
3767 if (queue->alloc <= queue->nr) {
3768 queue->alloc = alloc_nr(queue->alloc);
3769 queue->queue = xrealloc(queue->queue,
3770 sizeof(dp) * queue->alloc);
3771 }
3772 queue->queue[queue->nr++] = dp;
3773}
3774
3775struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
3776 struct diff_filespec *one,
3777 struct diff_filespec *two)
3778{
3779 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
3780 dp->one = one;
3781 dp->two = two;
3782 if (queue)
3783 diff_q(queue, dp);
3784 return dp;
3785}
3786
3787void diff_free_filepair(struct diff_filepair *p)
3788{
3789 free_filespec(p->one);
3790 free_filespec(p->two);
3791 free(p);
3792}
3793
3794/* This is different from find_unique_abbrev() in that
3795 * it stuffs the result with dots for alignment.
3796 */
3797const char *diff_unique_abbrev(const unsigned char *sha1, int len)
3798{
3799 int abblen;
3800 const char *abbrev;
3801 if (len == 40)
3802 return sha1_to_hex(sha1);
3803
3804 abbrev = find_unique_abbrev(sha1, len);
3805 abblen = strlen(abbrev);
3806 if (abblen < 37) {
3807 static char hex[41];
3808 if (len < abblen && abblen <= len + 2)
3809 sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
3810 else
3811 sprintf(hex, "%s...", abbrev);
3812 return hex;
3813 }
3814 return sha1_to_hex(sha1);
3815}
3816
3817static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
3818{
3819 int line_termination = opt->line_termination;
3820 int inter_name_termination = line_termination ? '\t' : '\0';
3821 if (opt->output_prefix) {
3822 struct strbuf *msg = NULL;
3823 msg = opt->output_prefix(opt, opt->output_prefix_data);
3824 fprintf(opt->file, "%s", msg->buf);
3825 }
3826
3827 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
3828 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
3829 diff_unique_abbrev(p->one->sha1, opt->abbrev));
3830 fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
3831 }
3832 if (p->score) {
3833 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
3834 inter_name_termination);
3835 } else {
3836 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
3837 }
3838
3839 if (p->status == DIFF_STATUS_COPIED ||
3840 p->status == DIFF_STATUS_RENAMED) {
3841 const char *name_a, *name_b;
3842 name_a = p->one->path;
3843 name_b = p->two->path;
3844 strip_prefix(opt->prefix_length, &name_a, &name_b);
3845 write_name_quoted(name_a, opt->file, inter_name_termination);
3846 write_name_quoted(name_b, opt->file, line_termination);
3847 } else {
3848 const char *name_a, *name_b;
3849 name_a = p->one->mode ? p->one->path : p->two->path;
3850 name_b = NULL;
3851 strip_prefix(opt->prefix_length, &name_a, &name_b);
3852 write_name_quoted(name_a, opt->file, line_termination);
3853 }
3854}
3855
3856int diff_unmodified_pair(struct diff_filepair *p)
3857{
3858 /* This function is written stricter than necessary to support
3859 * the currently implemented transformers, but the idea is to
3860 * let transformers to produce diff_filepairs any way they want,
3861 * and filter and clean them up here before producing the output.
3862 */
3863 struct diff_filespec *one = p->one, *two = p->two;
3864
3865 if (DIFF_PAIR_UNMERGED(p))
3866 return 0; /* unmerged is interesting */
3867
3868 /* deletion, addition, mode or type change
3869 * and rename are all interesting.
3870 */
3871 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
3872 DIFF_PAIR_MODE_CHANGED(p) ||
3873 strcmp(one->path, two->path))
3874 return 0;
3875
3876 /* both are valid and point at the same path. that is, we are
3877 * dealing with a change.
3878 */
3879 if (one->sha1_valid && two->sha1_valid &&
3880 !hashcmp(one->sha1, two->sha1) &&
3881 !one->dirty_submodule && !two->dirty_submodule)
3882 return 1; /* no change */
3883 if (!one->sha1_valid && !two->sha1_valid)
3884 return 1; /* both look at the same file on the filesystem. */
3885 return 0;
3886}
3887
3888static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
3889{
3890 if (diff_unmodified_pair(p))
3891 return;
3892
3893 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3894 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3895 return; /* no tree diffs in patch format */
3896
3897 run_diff(p, o);
3898}
3899
3900static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
3901 struct diffstat_t *diffstat)
3902{
3903 if (diff_unmodified_pair(p))
3904 return;
3905
3906 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3907 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3908 return; /* no useful stat for tree diffs */
3909
3910 run_diffstat(p, o, diffstat);
3911}
3912
3913static void diff_flush_checkdiff(struct diff_filepair *p,
3914 struct diff_options *o)
3915{
3916 if (diff_unmodified_pair(p))
3917 return;
3918
3919 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3920 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3921 return; /* nothing to check in tree diffs */
3922
3923 run_checkdiff(p, o);
3924}
3925
3926int diff_queue_is_empty(void)
3927{
3928 struct diff_queue_struct *q = &diff_queued_diff;
3929 int i;
3930 for (i = 0; i < q->nr; i++)
3931 if (!diff_unmodified_pair(q->queue[i]))
3932 return 0;
3933 return 1;
3934}
3935
3936#if DIFF_DEBUG
3937void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
3938{
3939 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
3940 x, one ? one : "",
3941 s->path,
3942 DIFF_FILE_VALID(s) ? "valid" : "invalid",
3943 s->mode,
3944 s->sha1_valid ? sha1_to_hex(s->sha1) : "");
3945 fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
3946 x, one ? one : "",
3947 s->size, s->xfrm_flags);
3948}
3949
3950void diff_debug_filepair(const struct diff_filepair *p, int i)
3951{
3952 diff_debug_filespec(p->one, i, "one");
3953 diff_debug_filespec(p->two, i, "two");
3954 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
3955 p->score, p->status ? p->status : '?',
3956 p->one->rename_used, p->broken_pair);
3957}
3958
3959void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
3960{
3961 int i;
3962 if (msg)
3963 fprintf(stderr, "%s\n", msg);
3964 fprintf(stderr, "q->nr = %d\n", q->nr);
3965 for (i = 0; i < q->nr; i++) {
3966 struct diff_filepair *p = q->queue[i];
3967 diff_debug_filepair(p, i);
3968 }
3969}
3970#endif
3971
3972static void diff_resolve_rename_copy(void)
3973{
3974 int i;
3975 struct diff_filepair *p;
3976 struct diff_queue_struct *q = &diff_queued_diff;
3977
3978 diff_debug_queue("resolve-rename-copy", q);
3979
3980 for (i = 0; i < q->nr; i++) {
3981 p = q->queue[i];
3982 p->status = 0; /* undecided */
3983 if (DIFF_PAIR_UNMERGED(p))
3984 p->status = DIFF_STATUS_UNMERGED;
3985 else if (!DIFF_FILE_VALID(p->one))
3986 p->status = DIFF_STATUS_ADDED;
3987 else if (!DIFF_FILE_VALID(p->two))
3988 p->status = DIFF_STATUS_DELETED;
3989 else if (DIFF_PAIR_TYPE_CHANGED(p))
3990 p->status = DIFF_STATUS_TYPE_CHANGED;
3991
3992 /* from this point on, we are dealing with a pair
3993 * whose both sides are valid and of the same type, i.e.
3994 * either in-place edit or rename/copy edit.
3995 */
3996 else if (DIFF_PAIR_RENAME(p)) {
3997 /*
3998 * A rename might have re-connected a broken
3999 * pair up, causing the pathnames to be the
4000 * same again. If so, that's not a rename at
4001 * all, just a modification..
4002 *
4003 * Otherwise, see if this source was used for
4004 * multiple renames, in which case we decrement
4005 * the count, and call it a copy.
4006 */
4007 if (!strcmp(p->one->path, p->two->path))
4008 p->status = DIFF_STATUS_MODIFIED;
4009 else if (--p->one->rename_used > 0)
4010 p->status = DIFF_STATUS_COPIED;
4011 else
4012 p->status = DIFF_STATUS_RENAMED;
4013 }
4014 else if (hashcmp(p->one->sha1, p->two->sha1) ||
4015 p->one->mode != p->two->mode ||
4016 p->one->dirty_submodule ||
4017 p->two->dirty_submodule ||
4018 is_null_sha1(p->one->sha1))
4019 p->status = DIFF_STATUS_MODIFIED;
4020 else {
4021 /* This is a "no-change" entry and should not
4022 * happen anymore, but prepare for broken callers.
4023 */
4024 error("feeding unmodified %s to diffcore",
4025 p->one->path);
4026 p->status = DIFF_STATUS_UNKNOWN;
4027 }
4028 }
4029 diff_debug_queue("resolve-rename-copy done", q);
4030}
4031
4032static int check_pair_status(struct diff_filepair *p)
4033{
4034 switch (p->status) {
4035 case DIFF_STATUS_UNKNOWN:
4036 return 0;
4037 case 0:
4038 die("internal error in diff-resolve-rename-copy");
4039 default:
4040 return 1;
4041 }
4042}
4043
4044static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4045{
4046 int fmt = opt->output_format;
4047
4048 if (fmt & DIFF_FORMAT_CHECKDIFF)
4049 diff_flush_checkdiff(p, opt);
4050 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4051 diff_flush_raw(p, opt);
4052 else if (fmt & DIFF_FORMAT_NAME) {
4053 const char *name_a, *name_b;
4054 name_a = p->two->path;
4055 name_b = NULL;
4056 strip_prefix(opt->prefix_length, &name_a, &name_b);
4057 write_name_quoted(name_a, opt->file, opt->line_termination);
4058 }
4059}
4060
4061static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4062{
4063 if (fs->mode)
4064 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4065 else
4066 fprintf(file, " %s ", newdelete);
4067 write_name_quoted(fs->path, file, '\n');
4068}
4069
4070
4071static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4072 const char *line_prefix)
4073{
4074 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4075 fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4076 p->two->mode, show_name ? ' ' : '\n');
4077 if (show_name) {
4078 write_name_quoted(p->two->path, file, '\n');
4079 }
4080 }
4081}
4082
4083static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4084 const char *line_prefix)
4085{
4086 char *names = pprint_rename(p->one->path, p->two->path);
4087
4088 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4089 free(names);
4090 show_mode_change(file, p, 0, line_prefix);
4091}
4092
4093static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4094{
4095 FILE *file = opt->file;
4096 char *line_prefix = "";
4097
4098 if (opt->output_prefix) {
4099 struct strbuf *buf = opt->output_prefix(opt, opt->output_prefix_data);
4100 line_prefix = buf->buf;
4101 }
4102
4103 switch(p->status) {
4104 case DIFF_STATUS_DELETED:
4105 fputs(line_prefix, file);
4106 show_file_mode_name(file, "delete", p->one);
4107 break;
4108 case DIFF_STATUS_ADDED:
4109 fputs(line_prefix, file);
4110 show_file_mode_name(file, "create", p->two);
4111 break;
4112 case DIFF_STATUS_COPIED:
4113 fputs(line_prefix, file);
4114 show_rename_copy(file, "copy", p, line_prefix);
4115 break;
4116 case DIFF_STATUS_RENAMED:
4117 fputs(line_prefix, file);
4118 show_rename_copy(file, "rename", p, line_prefix);
4119 break;
4120 default:
4121 if (p->score) {
4122 fprintf(file, "%s rewrite ", line_prefix);
4123 write_name_quoted(p->two->path, file, ' ');
4124 fprintf(file, "(%d%%)\n", similarity_index(p));
4125 }
4126 show_mode_change(file, p, !p->score, line_prefix);
4127 break;
4128 }
4129}
4130
4131struct patch_id_t {
4132 git_SHA_CTX *ctx;
4133 int patchlen;
4134};
4135
4136static int remove_space(char *line, int len)
4137{
4138 int i;
4139 char *dst = line;
4140 unsigned char c;
4141
4142 for (i = 0; i < len; i++)
4143 if (!isspace((c = line[i])))
4144 *dst++ = c;
4145
4146 return dst - line;
4147}
4148
4149static void patch_id_consume(void *priv, char *line, unsigned long len)
4150{
4151 struct patch_id_t *data = priv;
4152 int new_len;
4153
4154 /* Ignore line numbers when computing the SHA1 of the patch */
4155 if (!prefixcmp(line, "@@ -"))
4156 return;
4157
4158 new_len = remove_space(line, len);
4159
4160 git_SHA1_Update(data->ctx, line, new_len);
4161 data->patchlen += new_len;
4162}
4163
4164/* returns 0 upon success, and writes result into sha1 */
4165static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
4166{
4167 struct diff_queue_struct *q = &diff_queued_diff;
4168 int i;
4169 git_SHA_CTX ctx;
4170 struct patch_id_t data;
4171 char buffer[PATH_MAX * 4 + 20];
4172
4173 git_SHA1_Init(&ctx);
4174 memset(&data, 0, sizeof(struct patch_id_t));
4175 data.ctx = &ctx;
4176
4177 for (i = 0; i < q->nr; i++) {
4178 xpparam_t xpp;
4179 xdemitconf_t xecfg;
4180 mmfile_t mf1, mf2;
4181 struct diff_filepair *p = q->queue[i];
4182 int len1, len2;
4183
4184 memset(&xpp, 0, sizeof(xpp));
4185 memset(&xecfg, 0, sizeof(xecfg));
4186 if (p->status == 0)
4187 return error("internal diff status error");
4188 if (p->status == DIFF_STATUS_UNKNOWN)
4189 continue;
4190 if (diff_unmodified_pair(p))
4191 continue;
4192 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4193 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4194 continue;
4195 if (DIFF_PAIR_UNMERGED(p))
4196 continue;
4197
4198 diff_fill_sha1_info(p->one);
4199 diff_fill_sha1_info(p->two);
4200 if (fill_mmfile(&mf1, p->one) < 0 ||
4201 fill_mmfile(&mf2, p->two) < 0)
4202 return error("unable to read files to diff");
4203
4204 len1 = remove_space(p->one->path, strlen(p->one->path));
4205 len2 = remove_space(p->two->path, strlen(p->two->path));
4206 if (p->one->mode == 0)
4207 len1 = snprintf(buffer, sizeof(buffer),
4208 "diff--gita/%.*sb/%.*s"
4209 "newfilemode%06o"
4210 "---/dev/null"
4211 "+++b/%.*s",
4212 len1, p->one->path,
4213 len2, p->two->path,
4214 p->two->mode,
4215 len2, p->two->path);
4216 else if (p->two->mode == 0)
4217 len1 = snprintf(buffer, sizeof(buffer),
4218 "diff--gita/%.*sb/%.*s"
4219 "deletedfilemode%06o"
4220 "---a/%.*s"
4221 "+++/dev/null",
4222 len1, p->one->path,
4223 len2, p->two->path,
4224 p->one->mode,
4225 len1, p->one->path);
4226 else
4227 len1 = snprintf(buffer, sizeof(buffer),
4228 "diff--gita/%.*sb/%.*s"
4229 "---a/%.*s"
4230 "+++b/%.*s",
4231 len1, p->one->path,
4232 len2, p->two->path,
4233 len1, p->one->path,
4234 len2, p->two->path);
4235 git_SHA1_Update(&ctx, buffer, len1);
4236
4237 if (diff_filespec_is_binary(p->one) ||
4238 diff_filespec_is_binary(p->two)) {
4239 git_SHA1_Update(&ctx, sha1_to_hex(p->one->sha1), 40);
4240 git_SHA1_Update(&ctx, sha1_to_hex(p->two->sha1), 40);
4241 continue;
4242 }
4243
4244 xpp.flags = 0;
4245 xecfg.ctxlen = 3;
4246 xecfg.flags = 0;
4247 xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4248 &xpp, &xecfg);
4249 }
4250
4251 git_SHA1_Final(sha1, &ctx);
4252 return 0;
4253}
4254
4255int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
4256{
4257 struct diff_queue_struct *q = &diff_queued_diff;
4258 int i;
4259 int result = diff_get_patch_id(options, sha1);
4260
4261 for (i = 0; i < q->nr; i++)
4262 diff_free_filepair(q->queue[i]);
4263
4264 free(q->queue);
4265 DIFF_QUEUE_CLEAR(q);
4266
4267 return result;
4268}
4269
4270static int is_summary_empty(const struct diff_queue_struct *q)
4271{
4272 int i;
4273
4274 for (i = 0; i < q->nr; i++) {
4275 const struct diff_filepair *p = q->queue[i];
4276
4277 switch (p->status) {
4278 case DIFF_STATUS_DELETED:
4279 case DIFF_STATUS_ADDED:
4280 case DIFF_STATUS_COPIED:
4281 case DIFF_STATUS_RENAMED:
4282 return 0;
4283 default:
4284 if (p->score)
4285 return 0;
4286 if (p->one->mode && p->two->mode &&
4287 p->one->mode != p->two->mode)
4288 return 0;
4289 break;
4290 }
4291 }
4292 return 1;
4293}
4294
4295static const char rename_limit_warning[] =
4296"inexact rename detection was skipped due to too many files.";
4297
4298static const char degrade_cc_to_c_warning[] =
4299"only found copies from modified paths due to too many files.";
4300
4301static const char rename_limit_advice[] =
4302"you may want to set your %s variable to at least "
4303"%d and retry the command.";
4304
4305void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4306{
4307 if (degraded_cc)
4308 warning(degrade_cc_to_c_warning);
4309 else if (needed)
4310 warning(rename_limit_warning);
4311 else
4312 return;
4313 if (0 < needed && needed < 32767)
4314 warning(rename_limit_advice, varname, needed);
4315}
4316
4317void diff_flush(struct diff_options *options)
4318{
4319 struct diff_queue_struct *q = &diff_queued_diff;
4320 int i, output_format = options->output_format;
4321 int separator = 0;
4322 int dirstat_by_line = 0;
4323
4324 /*
4325 * Order: raw, stat, summary, patch
4326 * or: name/name-status/checkdiff (other bits clear)
4327 */
4328 if (!q->nr)
4329 goto free_queue;
4330
4331 if (output_format & (DIFF_FORMAT_RAW |
4332 DIFF_FORMAT_NAME |
4333 DIFF_FORMAT_NAME_STATUS |
4334 DIFF_FORMAT_CHECKDIFF)) {
4335 for (i = 0; i < q->nr; i++) {
4336 struct diff_filepair *p = q->queue[i];
4337 if (check_pair_status(p))
4338 flush_one_pair(p, options);
4339 }
4340 separator++;
4341 }
4342
4343 if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4344 dirstat_by_line = 1;
4345
4346 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4347 dirstat_by_line) {
4348 struct diffstat_t diffstat;
4349
4350 memset(&diffstat, 0, sizeof(struct diffstat_t));
4351 for (i = 0; i < q->nr; i++) {
4352 struct diff_filepair *p = q->queue[i];
4353 if (check_pair_status(p))
4354 diff_flush_stat(p, options, &diffstat);
4355 }
4356 if (output_format & DIFF_FORMAT_NUMSTAT)
4357 show_numstat(&diffstat, options);
4358 if (output_format & DIFF_FORMAT_DIFFSTAT)
4359 show_stats(&diffstat, options);
4360 if (output_format & DIFF_FORMAT_SHORTSTAT)
4361 show_shortstats(&diffstat, options);
4362 if (output_format & DIFF_FORMAT_DIRSTAT)
4363 show_dirstat_by_line(&diffstat, options);
4364 free_diffstat_info(&diffstat);
4365 separator++;
4366 }
4367 if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4368 show_dirstat(options);
4369
4370 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4371 for (i = 0; i < q->nr; i++) {
4372 diff_summary(options, q->queue[i]);
4373 }
4374 separator++;
4375 }
4376
4377 if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4378 DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4379 DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4380 /*
4381 * run diff_flush_patch for the exit status. setting
4382 * options->file to /dev/null should be safe, becaue we
4383 * aren't supposed to produce any output anyway.
4384 */
4385 if (options->close_file)
4386 fclose(options->file);
4387 options->file = fopen("/dev/null", "w");
4388 if (!options->file)
4389 die_errno("Could not open /dev/null");
4390 options->close_file = 1;
4391 for (i = 0; i < q->nr; i++) {
4392 struct diff_filepair *p = q->queue[i];
4393 if (check_pair_status(p))
4394 diff_flush_patch(p, options);
4395 if (options->found_changes)
4396 break;
4397 }
4398 }
4399
4400 if (output_format & DIFF_FORMAT_PATCH) {
4401 if (separator) {
4402 putc(options->line_termination, options->file);
4403 if (options->stat_sep) {
4404 /* attach patch instead of inline */
4405 fputs(options->stat_sep, options->file);
4406 }
4407 }
4408
4409 for (i = 0; i < q->nr; i++) {
4410 struct diff_filepair *p = q->queue[i];
4411 if (check_pair_status(p))
4412 diff_flush_patch(p, options);
4413 }
4414 }
4415
4416 if (output_format & DIFF_FORMAT_CALLBACK)
4417 options->format_callback(q, options, options->format_callback_data);
4418
4419 for (i = 0; i < q->nr; i++)
4420 diff_free_filepair(q->queue[i]);
4421free_queue:
4422 free(q->queue);
4423 DIFF_QUEUE_CLEAR(q);
4424 if (options->close_file)
4425 fclose(options->file);
4426
4427 /*
4428 * Report the content-level differences with HAS_CHANGES;
4429 * diff_addremove/diff_change does not set the bit when
4430 * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4431 */
4432 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4433 if (options->found_changes)
4434 DIFF_OPT_SET(options, HAS_CHANGES);
4435 else
4436 DIFF_OPT_CLR(options, HAS_CHANGES);
4437 }
4438}
4439
4440static void diffcore_apply_filter(const char *filter)
4441{
4442 int i;
4443 struct diff_queue_struct *q = &diff_queued_diff;
4444 struct diff_queue_struct outq;
4445 DIFF_QUEUE_CLEAR(&outq);
4446
4447 if (!filter)
4448 return;
4449
4450 if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
4451 int found;
4452 for (i = found = 0; !found && i < q->nr; i++) {
4453 struct diff_filepair *p = q->queue[i];
4454 if (((p->status == DIFF_STATUS_MODIFIED) &&
4455 ((p->score &&
4456 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4457 (!p->score &&
4458 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4459 ((p->status != DIFF_STATUS_MODIFIED) &&
4460 strchr(filter, p->status)))
4461 found++;
4462 }
4463 if (found)
4464 return;
4465
4466 /* otherwise we will clear the whole queue
4467 * by copying the empty outq at the end of this
4468 * function, but first clear the current entries
4469 * in the queue.
4470 */
4471 for (i = 0; i < q->nr; i++)
4472 diff_free_filepair(q->queue[i]);
4473 }
4474 else {
4475 /* Only the matching ones */
4476 for (i = 0; i < q->nr; i++) {
4477 struct diff_filepair *p = q->queue[i];
4478
4479 if (((p->status == DIFF_STATUS_MODIFIED) &&
4480 ((p->score &&
4481 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4482 (!p->score &&
4483 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4484 ((p->status != DIFF_STATUS_MODIFIED) &&
4485 strchr(filter, p->status)))
4486 diff_q(&outq, p);
4487 else
4488 diff_free_filepair(p);
4489 }
4490 }
4491 free(q->queue);
4492 *q = outq;
4493}
4494
4495/* Check whether two filespecs with the same mode and size are identical */
4496static int diff_filespec_is_identical(struct diff_filespec *one,
4497 struct diff_filespec *two)
4498{
4499 if (S_ISGITLINK(one->mode))
4500 return 0;
4501 if (diff_populate_filespec(one, 0))
4502 return 0;
4503 if (diff_populate_filespec(two, 0))
4504 return 0;
4505 return !memcmp(one->data, two->data, one->size);
4506}
4507
4508static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4509{
4510 int i;
4511 struct diff_queue_struct *q = &diff_queued_diff;
4512 struct diff_queue_struct outq;
4513 DIFF_QUEUE_CLEAR(&outq);
4514
4515 for (i = 0; i < q->nr; i++) {
4516 struct diff_filepair *p = q->queue[i];
4517
4518 /*
4519 * 1. Entries that come from stat info dirtiness
4520 * always have both sides (iow, not create/delete),
4521 * one side of the object name is unknown, with
4522 * the same mode and size. Keep the ones that
4523 * do not match these criteria. They have real
4524 * differences.
4525 *
4526 * 2. At this point, the file is known to be modified,
4527 * with the same mode and size, and the object
4528 * name of one side is unknown. Need to inspect
4529 * the identical contents.
4530 */
4531 if (!DIFF_FILE_VALID(p->one) || /* (1) */
4532 !DIFF_FILE_VALID(p->two) ||
4533 (p->one->sha1_valid && p->two->sha1_valid) ||
4534 (p->one->mode != p->two->mode) ||
4535 diff_populate_filespec(p->one, 1) ||
4536 diff_populate_filespec(p->two, 1) ||
4537 (p->one->size != p->two->size) ||
4538 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4539 diff_q(&outq, p);
4540 else {
4541 /*
4542 * The caller can subtract 1 from skip_stat_unmatch
4543 * to determine how many paths were dirty only
4544 * due to stat info mismatch.
4545 */
4546 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4547 diffopt->skip_stat_unmatch++;
4548 diff_free_filepair(p);
4549 }
4550 }
4551 free(q->queue);
4552 *q = outq;
4553}
4554
4555static int diffnamecmp(const void *a_, const void *b_)
4556{
4557 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4558 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4559 const char *name_a, *name_b;
4560
4561 name_a = a->one ? a->one->path : a->two->path;
4562 name_b = b->one ? b->one->path : b->two->path;
4563 return strcmp(name_a, name_b);
4564}
4565
4566void diffcore_fix_diff_index(struct diff_options *options)
4567{
4568 struct diff_queue_struct *q = &diff_queued_diff;
4569 qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4570}
4571
4572void diffcore_std(struct diff_options *options)
4573{
4574 if (options->skip_stat_unmatch)
4575 diffcore_skip_stat_unmatch(options);
4576 if (!options->found_follow) {
4577 /* See try_to_follow_renames() in tree-diff.c */
4578 if (options->break_opt != -1)
4579 diffcore_break(options->break_opt);
4580 if (options->detect_rename)
4581 diffcore_rename(options);
4582 if (options->break_opt != -1)
4583 diffcore_merge_broken();
4584 }
4585 if (options->pickaxe)
4586 diffcore_pickaxe(options);
4587 if (options->orderfile)
4588 diffcore_order(options->orderfile);
4589 if (!options->found_follow)
4590 /* See try_to_follow_renames() in tree-diff.c */
4591 diff_resolve_rename_copy();
4592 diffcore_apply_filter(options->filter);
4593
4594 if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4595 DIFF_OPT_SET(options, HAS_CHANGES);
4596 else
4597 DIFF_OPT_CLR(options, HAS_CHANGES);
4598
4599 options->found_follow = 0;
4600}
4601
4602int diff_result_code(struct diff_options *opt, int status)
4603{
4604 int result = 0;
4605
4606 diff_warn_rename_limit("diff.renamelimit",
4607 opt->needed_rename_limit,
4608 opt->degraded_cc_to_c);
4609 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4610 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4611 return status;
4612 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4613 DIFF_OPT_TST(opt, HAS_CHANGES))
4614 result |= 01;
4615 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4616 DIFF_OPT_TST(opt, CHECK_FAILED))
4617 result |= 02;
4618 return result;
4619}
4620
4621int diff_can_quit_early(struct diff_options *opt)
4622{
4623 return (DIFF_OPT_TST(opt, QUICK) &&
4624 !opt->filter &&
4625 DIFF_OPT_TST(opt, HAS_CHANGES));
4626}
4627
4628/*
4629 * Shall changes to this submodule be ignored?
4630 *
4631 * Submodule changes can be configured to be ignored separately for each path,
4632 * but that configuration can be overridden from the command line.
4633 */
4634static int is_submodule_ignored(const char *path, struct diff_options *options)
4635{
4636 int ignored = 0;
4637 unsigned orig_flags = options->flags;
4638 if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4639 set_diffopt_flags_from_submodule_config(options, path);
4640 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4641 ignored = 1;
4642 options->flags = orig_flags;
4643 return ignored;
4644}
4645
4646void diff_addremove(struct diff_options *options,
4647 int addremove, unsigned mode,
4648 const unsigned char *sha1,
4649 const char *concatpath, unsigned dirty_submodule)
4650{
4651 struct diff_filespec *one, *two;
4652
4653 if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
4654 return;
4655
4656 /* This may look odd, but it is a preparation for
4657 * feeding "there are unchanged files which should
4658 * not produce diffs, but when you are doing copy
4659 * detection you would need them, so here they are"
4660 * entries to the diff-core. They will be prefixed
4661 * with something like '=' or '*' (I haven't decided
4662 * which but should not make any difference).
4663 * Feeding the same new and old to diff_change()
4664 * also has the same effect.
4665 * Before the final output happens, they are pruned after
4666 * merged into rename/copy pairs as appropriate.
4667 */
4668 if (DIFF_OPT_TST(options, REVERSE_DIFF))
4669 addremove = (addremove == '+' ? '-' :
4670 addremove == '-' ? '+' : addremove);
4671
4672 if (options->prefix &&
4673 strncmp(concatpath, options->prefix, options->prefix_length))
4674 return;
4675
4676 one = alloc_filespec(concatpath);
4677 two = alloc_filespec(concatpath);
4678
4679 if (addremove != '+')
4680 fill_filespec(one, sha1, mode);
4681 if (addremove != '-') {
4682 fill_filespec(two, sha1, mode);
4683 two->dirty_submodule = dirty_submodule;
4684 }
4685
4686 diff_queue(&diff_queued_diff, one, two);
4687 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4688 DIFF_OPT_SET(options, HAS_CHANGES);
4689}
4690
4691void diff_change(struct diff_options *options,
4692 unsigned old_mode, unsigned new_mode,
4693 const unsigned char *old_sha1,
4694 const unsigned char *new_sha1,
4695 const char *concatpath,
4696 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
4697{
4698 struct diff_filespec *one, *two;
4699
4700 if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
4701 is_submodule_ignored(concatpath, options))
4702 return;
4703
4704 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
4705 unsigned tmp;
4706 const unsigned char *tmp_c;
4707 tmp = old_mode; old_mode = new_mode; new_mode = tmp;
4708 tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
4709 tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
4710 new_dirty_submodule = tmp;
4711 }
4712
4713 if (options->prefix &&
4714 strncmp(concatpath, options->prefix, options->prefix_length))
4715 return;
4716
4717 one = alloc_filespec(concatpath);
4718 two = alloc_filespec(concatpath);
4719 fill_filespec(one, old_sha1, old_mode);
4720 fill_filespec(two, new_sha1, new_mode);
4721 one->dirty_submodule = old_dirty_submodule;
4722 two->dirty_submodule = new_dirty_submodule;
4723
4724 diff_queue(&diff_queued_diff, one, two);
4725 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4726 DIFF_OPT_SET(options, HAS_CHANGES);
4727}
4728
4729struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
4730{
4731 struct diff_filepair *pair;
4732 struct diff_filespec *one, *two;
4733
4734 if (options->prefix &&
4735 strncmp(path, options->prefix, options->prefix_length))
4736 return NULL;
4737
4738 one = alloc_filespec(path);
4739 two = alloc_filespec(path);
4740 pair = diff_queue(&diff_queued_diff, one, two);
4741 pair->is_unmerged = 1;
4742 return pair;
4743}
4744
4745static char *run_textconv(const char *pgm, struct diff_filespec *spec,
4746 size_t *outsize)
4747{
4748 struct diff_tempfile *temp;
4749 const char *argv[3];
4750 const char **arg = argv;
4751 struct child_process child;
4752 struct strbuf buf = STRBUF_INIT;
4753 int err = 0;
4754
4755 temp = prepare_temp_file(spec->path, spec);
4756 *arg++ = pgm;
4757 *arg++ = temp->name;
4758 *arg = NULL;
4759
4760 memset(&child, 0, sizeof(child));
4761 child.use_shell = 1;
4762 child.argv = argv;
4763 child.out = -1;
4764 if (start_command(&child)) {
4765 remove_tempfile();
4766 return NULL;
4767 }
4768
4769 if (strbuf_read(&buf, child.out, 0) < 0)
4770 err = error("error reading from textconv command '%s'", pgm);
4771 close(child.out);
4772
4773 if (finish_command(&child) || err) {
4774 strbuf_release(&buf);
4775 remove_tempfile();
4776 return NULL;
4777 }
4778 remove_tempfile();
4779
4780 return strbuf_detach(&buf, outsize);
4781}
4782
4783size_t fill_textconv(struct userdiff_driver *driver,
4784 struct diff_filespec *df,
4785 char **outbuf)
4786{
4787 size_t size;
4788
4789 if (!driver || !driver->textconv) {
4790 if (!DIFF_FILE_VALID(df)) {
4791 *outbuf = "";
4792 return 0;
4793 }
4794 if (diff_populate_filespec(df, 0))
4795 die("unable to read files to diff");
4796 *outbuf = df->data;
4797 return df->size;
4798 }
4799
4800 if (driver->textconv_cache && df->sha1_valid) {
4801 *outbuf = notes_cache_get(driver->textconv_cache, df->sha1,
4802 &size);
4803 if (*outbuf)
4804 return size;
4805 }
4806
4807 *outbuf = run_textconv(driver->textconv, df, &size);
4808 if (!*outbuf)
4809 die("unable to read files to diff");
4810
4811 if (driver->textconv_cache && df->sha1_valid) {
4812 /* ignore errors, as we might be in a readonly repository */
4813 notes_cache_put(driver->textconv_cache, df->sha1, *outbuf,
4814 size);
4815 /*
4816 * we could save up changes and flush them all at the end,
4817 * but we would need an extra call after all diffing is done.
4818 * Since generating a cache entry is the slow path anyway,
4819 * this extra overhead probably isn't a big deal.
4820 */
4821 notes_cache_write(driver->textconv_cache);
4822 }
4823
4824 return size;
4825}