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 diff_filespec_load_driver(struct diff_filespec *one)
993{
994 /* Use already-loaded driver */
995 if (one->driver)
996 return;
997
998 if (S_ISREG(one->mode))
999 one->driver = userdiff_find_by_path(one->path);
1000
1001 /* Fallback to default settings */
1002 if (!one->driver)
1003 one->driver = userdiff_find_by_name("default");
1004}
1005
1006static const char *userdiff_word_regex(struct diff_filespec *one)
1007{
1008 diff_filespec_load_driver(one);
1009 return one->driver->word_regex;
1010}
1011
1012static void init_diff_words_data(struct emit_callback *ecbdata,
1013 struct diff_options *orig_opts,
1014 struct diff_filespec *one,
1015 struct diff_filespec *two)
1016{
1017 int i;
1018 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1019 memcpy(o, orig_opts, sizeof(struct diff_options));
1020
1021 ecbdata->diff_words =
1022 xcalloc(1, sizeof(struct diff_words_data));
1023 ecbdata->diff_words->type = o->word_diff;
1024 ecbdata->diff_words->opt = o;
1025 if (!o->word_regex)
1026 o->word_regex = userdiff_word_regex(one);
1027 if (!o->word_regex)
1028 o->word_regex = userdiff_word_regex(two);
1029 if (!o->word_regex)
1030 o->word_regex = diff_word_regex_cfg;
1031 if (o->word_regex) {
1032 ecbdata->diff_words->word_regex = (regex_t *)
1033 xmalloc(sizeof(regex_t));
1034 if (regcomp(ecbdata->diff_words->word_regex,
1035 o->word_regex,
1036 REG_EXTENDED | REG_NEWLINE))
1037 die ("Invalid regular expression: %s",
1038 o->word_regex);
1039 }
1040 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1041 if (o->word_diff == diff_words_styles[i].type) {
1042 ecbdata->diff_words->style =
1043 &diff_words_styles[i];
1044 break;
1045 }
1046 }
1047 if (want_color(o->use_color)) {
1048 struct diff_words_style *st = ecbdata->diff_words->style;
1049 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1050 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1051 st->ctx.color = diff_get_color_opt(o, DIFF_PLAIN);
1052 }
1053}
1054
1055static void free_diff_words_data(struct emit_callback *ecbdata)
1056{
1057 if (ecbdata->diff_words) {
1058 diff_words_flush(ecbdata);
1059 free (ecbdata->diff_words->opt);
1060 free (ecbdata->diff_words->minus.text.ptr);
1061 free (ecbdata->diff_words->minus.orig);
1062 free (ecbdata->diff_words->plus.text.ptr);
1063 free (ecbdata->diff_words->plus.orig);
1064 if (ecbdata->diff_words->word_regex) {
1065 regfree(ecbdata->diff_words->word_regex);
1066 free(ecbdata->diff_words->word_regex);
1067 }
1068 free(ecbdata->diff_words);
1069 ecbdata->diff_words = NULL;
1070 }
1071}
1072
1073const char *diff_get_color(int diff_use_color, enum color_diff ix)
1074{
1075 if (want_color(diff_use_color))
1076 return diff_colors[ix];
1077 return "";
1078}
1079
1080static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1081{
1082 const char *cp;
1083 unsigned long allot;
1084 size_t l = len;
1085
1086 if (ecb->truncate)
1087 return ecb->truncate(line, len);
1088 cp = line;
1089 allot = l;
1090 while (0 < l) {
1091 (void) utf8_width(&cp, &l);
1092 if (!cp)
1093 break; /* truncated in the middle? */
1094 }
1095 return allot - l;
1096}
1097
1098static void find_lno(const char *line, struct emit_callback *ecbdata)
1099{
1100 const char *p;
1101 ecbdata->lno_in_preimage = 0;
1102 ecbdata->lno_in_postimage = 0;
1103 p = strchr(line, '-');
1104 if (!p)
1105 return; /* cannot happen */
1106 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1107 p = strchr(p, '+');
1108 if (!p)
1109 return; /* cannot happen */
1110 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1111}
1112
1113static void fn_out_consume(void *priv, char *line, unsigned long len)
1114{
1115 struct emit_callback *ecbdata = priv;
1116 const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1117 const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
1118 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1119 struct diff_options *o = ecbdata->opt;
1120 char *line_prefix = "";
1121 struct strbuf *msgbuf;
1122
1123 if (o && o->output_prefix) {
1124 msgbuf = o->output_prefix(o, o->output_prefix_data);
1125 line_prefix = msgbuf->buf;
1126 }
1127
1128 if (ecbdata->header) {
1129 fprintf(ecbdata->opt->file, "%s", ecbdata->header->buf);
1130 strbuf_reset(ecbdata->header);
1131 ecbdata->header = NULL;
1132 }
1133 *(ecbdata->found_changesp) = 1;
1134
1135 if (ecbdata->label_path[0]) {
1136 const char *name_a_tab, *name_b_tab;
1137
1138 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1139 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1140
1141 fprintf(ecbdata->opt->file, "%s%s--- %s%s%s\n",
1142 line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1143 fprintf(ecbdata->opt->file, "%s%s+++ %s%s%s\n",
1144 line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1145 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1146 }
1147
1148 if (diff_suppress_blank_empty
1149 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1150 line[0] = '\n';
1151 len = 1;
1152 }
1153
1154 if (line[0] == '@') {
1155 if (ecbdata->diff_words)
1156 diff_words_flush(ecbdata);
1157 len = sane_truncate_line(ecbdata, line, len);
1158 find_lno(line, ecbdata);
1159 emit_hunk_header(ecbdata, line, len);
1160 if (line[len-1] != '\n')
1161 putc('\n', ecbdata->opt->file);
1162 return;
1163 }
1164
1165 if (len < 1) {
1166 emit_line(ecbdata->opt, reset, reset, line, len);
1167 if (ecbdata->diff_words
1168 && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1169 fputs("~\n", ecbdata->opt->file);
1170 return;
1171 }
1172
1173 if (ecbdata->diff_words) {
1174 if (line[0] == '-') {
1175 diff_words_append(line, len,
1176 &ecbdata->diff_words->minus);
1177 return;
1178 } else if (line[0] == '+') {
1179 diff_words_append(line, len,
1180 &ecbdata->diff_words->plus);
1181 return;
1182 } else if (!prefixcmp(line, "\\ ")) {
1183 /*
1184 * Eat the "no newline at eof" marker as if we
1185 * saw a "+" or "-" line with nothing on it,
1186 * and return without diff_words_flush() to
1187 * defer processing. If this is the end of
1188 * preimage, more "+" lines may come after it.
1189 */
1190 return;
1191 }
1192 diff_words_flush(ecbdata);
1193 if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1194 emit_line(ecbdata->opt, plain, reset, line, len);
1195 fputs("~\n", ecbdata->opt->file);
1196 } else {
1197 /*
1198 * Skip the prefix character, if any. With
1199 * diff_suppress_blank_empty, there may be
1200 * none.
1201 */
1202 if (line[0] != '\n') {
1203 line++;
1204 len--;
1205 }
1206 emit_line(ecbdata->opt, plain, reset, line, len);
1207 }
1208 return;
1209 }
1210
1211 if (line[0] != '+') {
1212 const char *color =
1213 diff_get_color(ecbdata->color_diff,
1214 line[0] == '-' ? DIFF_FILE_OLD : DIFF_PLAIN);
1215 ecbdata->lno_in_preimage++;
1216 if (line[0] == ' ')
1217 ecbdata->lno_in_postimage++;
1218 emit_line(ecbdata->opt, color, reset, line, len);
1219 } else {
1220 ecbdata->lno_in_postimage++;
1221 emit_add_line(reset, ecbdata, line + 1, len - 1);
1222 }
1223}
1224
1225static char *pprint_rename(const char *a, const char *b)
1226{
1227 const char *old = a;
1228 const char *new = b;
1229 struct strbuf name = STRBUF_INIT;
1230 int pfx_length, sfx_length;
1231 int len_a = strlen(a);
1232 int len_b = strlen(b);
1233 int a_midlen, b_midlen;
1234 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1235 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1236
1237 if (qlen_a || qlen_b) {
1238 quote_c_style(a, &name, NULL, 0);
1239 strbuf_addstr(&name, " => ");
1240 quote_c_style(b, &name, NULL, 0);
1241 return strbuf_detach(&name, NULL);
1242 }
1243
1244 /* Find common prefix */
1245 pfx_length = 0;
1246 while (*old && *new && *old == *new) {
1247 if (*old == '/')
1248 pfx_length = old - a + 1;
1249 old++;
1250 new++;
1251 }
1252
1253 /* Find common suffix */
1254 old = a + len_a;
1255 new = b + len_b;
1256 sfx_length = 0;
1257 while (a <= old && b <= new && *old == *new) {
1258 if (*old == '/')
1259 sfx_length = len_a - (old - a);
1260 old--;
1261 new--;
1262 }
1263
1264 /*
1265 * pfx{mid-a => mid-b}sfx
1266 * {pfx-a => pfx-b}sfx
1267 * pfx{sfx-a => sfx-b}
1268 * name-a => name-b
1269 */
1270 a_midlen = len_a - pfx_length - sfx_length;
1271 b_midlen = len_b - pfx_length - sfx_length;
1272 if (a_midlen < 0)
1273 a_midlen = 0;
1274 if (b_midlen < 0)
1275 b_midlen = 0;
1276
1277 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1278 if (pfx_length + sfx_length) {
1279 strbuf_add(&name, a, pfx_length);
1280 strbuf_addch(&name, '{');
1281 }
1282 strbuf_add(&name, a + pfx_length, a_midlen);
1283 strbuf_addstr(&name, " => ");
1284 strbuf_add(&name, b + pfx_length, b_midlen);
1285 if (pfx_length + sfx_length) {
1286 strbuf_addch(&name, '}');
1287 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1288 }
1289 return strbuf_detach(&name, NULL);
1290}
1291
1292struct diffstat_t {
1293 int nr;
1294 int alloc;
1295 struct diffstat_file {
1296 char *from_name;
1297 char *name;
1298 char *print_name;
1299 unsigned is_unmerged:1;
1300 unsigned is_binary:1;
1301 unsigned is_renamed:1;
1302 uintmax_t added, deleted;
1303 } **files;
1304};
1305
1306static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1307 const char *name_a,
1308 const char *name_b)
1309{
1310 struct diffstat_file *x;
1311 x = xcalloc(sizeof (*x), 1);
1312 if (diffstat->nr == diffstat->alloc) {
1313 diffstat->alloc = alloc_nr(diffstat->alloc);
1314 diffstat->files = xrealloc(diffstat->files,
1315 diffstat->alloc * sizeof(x));
1316 }
1317 diffstat->files[diffstat->nr++] = x;
1318 if (name_b) {
1319 x->from_name = xstrdup(name_a);
1320 x->name = xstrdup(name_b);
1321 x->is_renamed = 1;
1322 }
1323 else {
1324 x->from_name = NULL;
1325 x->name = xstrdup(name_a);
1326 }
1327 return x;
1328}
1329
1330static void diffstat_consume(void *priv, char *line, unsigned long len)
1331{
1332 struct diffstat_t *diffstat = priv;
1333 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1334
1335 if (line[0] == '+')
1336 x->added++;
1337 else if (line[0] == '-')
1338 x->deleted++;
1339}
1340
1341const char mime_boundary_leader[] = "------------";
1342
1343static int scale_linear(int it, int width, int max_change)
1344{
1345 if (!it)
1346 return 0;
1347 /*
1348 * make sure that at least one '-' or '+' is printed if
1349 * there is any change to this path. The easiest way is to
1350 * scale linearly as if the alloted width is one column shorter
1351 * than it is, and then add 1 to the result.
1352 */
1353 return 1 + (it * (width - 1) / max_change);
1354}
1355
1356static void show_name(FILE *file,
1357 const char *prefix, const char *name, int len)
1358{
1359 fprintf(file, " %s%-*s |", prefix, len, name);
1360}
1361
1362static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1363{
1364 if (cnt <= 0)
1365 return;
1366 fprintf(file, "%s", set);
1367 while (cnt--)
1368 putc(ch, file);
1369 fprintf(file, "%s", reset);
1370}
1371
1372static void fill_print_name(struct diffstat_file *file)
1373{
1374 char *pname;
1375
1376 if (file->print_name)
1377 return;
1378
1379 if (!file->is_renamed) {
1380 struct strbuf buf = STRBUF_INIT;
1381 if (quote_c_style(file->name, &buf, NULL, 0)) {
1382 pname = strbuf_detach(&buf, NULL);
1383 } else {
1384 pname = file->name;
1385 strbuf_release(&buf);
1386 }
1387 } else {
1388 pname = pprint_rename(file->from_name, file->name);
1389 }
1390 file->print_name = pname;
1391}
1392
1393int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1394{
1395 struct strbuf sb = STRBUF_INIT;
1396 int ret;
1397
1398 if (!files) {
1399 assert(insertions == 0 && deletions == 0);
1400 return fputs(_(" 0 files changed\n"), fp);
1401 }
1402
1403 strbuf_addf(&sb,
1404 Q_(" %d file changed", " %d files changed", files),
1405 files);
1406
1407 /*
1408 * For binary diff, the caller may want to print "x files
1409 * changed" with insertions == 0 && deletions == 0.
1410 *
1411 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1412 * is probably less confusing (i.e skip over "2 files changed
1413 * but nothing about added/removed lines? Is this a bug in Git?").
1414 */
1415 if (insertions || deletions == 0) {
1416 /*
1417 * TRANSLATORS: "+" in (+) is a line addition marker;
1418 * do not translate it.
1419 */
1420 strbuf_addf(&sb,
1421 Q_(", %d insertion(+)", ", %d insertions(+)",
1422 insertions),
1423 insertions);
1424 }
1425
1426 if (deletions || insertions == 0) {
1427 /*
1428 * TRANSLATORS: "-" in (-) is a line removal marker;
1429 * do not translate it.
1430 */
1431 strbuf_addf(&sb,
1432 Q_(", %d deletion(-)", ", %d deletions(-)",
1433 deletions),
1434 deletions);
1435 }
1436 strbuf_addch(&sb, '\n');
1437 ret = fputs(sb.buf, fp);
1438 strbuf_release(&sb);
1439 return ret;
1440}
1441
1442static void show_stats(struct diffstat_t *data, struct diff_options *options)
1443{
1444 int i, len, add, del, adds = 0, dels = 0;
1445 uintmax_t max_change = 0, max_len = 0;
1446 int total_files = data->nr, count;
1447 int width, name_width, graph_width, number_width = 0, bin_width = 0;
1448 const char *reset, *add_c, *del_c;
1449 const char *line_prefix = "";
1450 int extra_shown = 0;
1451 struct strbuf *msg = NULL;
1452
1453 if (data->nr == 0)
1454 return;
1455
1456 if (options->output_prefix) {
1457 msg = options->output_prefix(options, options->output_prefix_data);
1458 line_prefix = msg->buf;
1459 }
1460
1461 count = options->stat_count ? options->stat_count : data->nr;
1462
1463 reset = diff_get_color_opt(options, DIFF_RESET);
1464 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1465 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1466
1467 /*
1468 * Find the longest filename and max number of changes
1469 */
1470 for (i = 0; (i < count) && (i < data->nr); i++) {
1471 struct diffstat_file *file = data->files[i];
1472 uintmax_t change = file->added + file->deleted;
1473 if (!data->files[i]->is_renamed &&
1474 (change == 0)) {
1475 count++; /* not shown == room for one more */
1476 continue;
1477 }
1478 fill_print_name(file);
1479 len = strlen(file->print_name);
1480 if (max_len < len)
1481 max_len = len;
1482
1483 if (file->is_unmerged) {
1484 /* "Unmerged" is 8 characters */
1485 bin_width = bin_width < 8 ? 8 : bin_width;
1486 continue;
1487 }
1488 if (file->is_binary) {
1489 /* "Bin XXX -> YYY bytes" */
1490 int w = 14 + decimal_width(file->added)
1491 + decimal_width(file->deleted);
1492 bin_width = bin_width < w ? w : bin_width;
1493 /* Display change counts aligned with "Bin" */
1494 number_width = 3;
1495 continue;
1496 }
1497
1498 if (max_change < change)
1499 max_change = change;
1500 }
1501 count = i; /* min(count, data->nr) */
1502
1503 /*
1504 * We have width = stat_width or term_columns() columns total.
1505 * We want a maximum of min(max_len, stat_name_width) for the name part.
1506 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1507 * We also need 1 for " " and 4 + decimal_width(max_change)
1508 * for " | NNNN " and one the empty column at the end, altogether
1509 * 6 + decimal_width(max_change).
1510 *
1511 * If there's not enough space, we will use the smaller of
1512 * stat_name_width (if set) and 5/8*width for the filename,
1513 * and the rest for constant elements + graph part, but no more
1514 * than stat_graph_width for the graph part.
1515 * (5/8 gives 50 for filename and 30 for the constant parts + graph
1516 * for the standard terminal size).
1517 *
1518 * In other words: stat_width limits the maximum width, and
1519 * stat_name_width fixes the maximum width of the filename,
1520 * and is also used to divide available columns if there
1521 * aren't enough.
1522 *
1523 * Binary files are displayed with "Bin XXX -> YYY bytes"
1524 * instead of the change count and graph. This part is treated
1525 * similarly to the graph part, except that it is not
1526 * "scaled". If total width is too small to accomodate the
1527 * guaranteed minimum width of the filename part and the
1528 * separators and this message, this message will "overflow"
1529 * making the line longer than the maximum width.
1530 */
1531
1532 if (options->stat_width == -1)
1533 width = term_columns() - options->output_prefix_length;
1534 else
1535 width = options->stat_width ? options->stat_width : 80;
1536 number_width = decimal_width(max_change) > number_width ?
1537 decimal_width(max_change) : number_width;
1538
1539 if (options->stat_graph_width == -1)
1540 options->stat_graph_width = diff_stat_graph_width;
1541
1542 /*
1543 * Guarantee 3/8*16==6 for the graph part
1544 * and 5/8*16==10 for the filename part
1545 */
1546 if (width < 16 + 6 + number_width)
1547 width = 16 + 6 + number_width;
1548
1549 /*
1550 * First assign sizes that are wanted, ignoring available width.
1551 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1552 * starting from "XXX" should fit in graph_width.
1553 */
1554 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1555 if (options->stat_graph_width &&
1556 options->stat_graph_width < graph_width)
1557 graph_width = options->stat_graph_width;
1558
1559 name_width = (options->stat_name_width > 0 &&
1560 options->stat_name_width < max_len) ?
1561 options->stat_name_width : max_len;
1562
1563 /*
1564 * Adjust adjustable widths not to exceed maximum width
1565 */
1566 if (name_width + number_width + 6 + graph_width > width) {
1567 if (graph_width > width * 3/8 - number_width - 6) {
1568 graph_width = width * 3/8 - number_width - 6;
1569 if (graph_width < 6)
1570 graph_width = 6;
1571 }
1572
1573 if (options->stat_graph_width &&
1574 graph_width > options->stat_graph_width)
1575 graph_width = options->stat_graph_width;
1576 if (name_width > width - number_width - 6 - graph_width)
1577 name_width = width - number_width - 6 - graph_width;
1578 else
1579 graph_width = width - number_width - 6 - name_width;
1580 }
1581
1582 /*
1583 * From here name_width is the width of the name area,
1584 * and graph_width is the width of the graph area.
1585 * max_change is used to scale graph properly.
1586 */
1587 for (i = 0; i < count; i++) {
1588 const char *prefix = "";
1589 char *name = data->files[i]->print_name;
1590 uintmax_t added = data->files[i]->added;
1591 uintmax_t deleted = data->files[i]->deleted;
1592 int name_len;
1593
1594 if (!data->files[i]->is_renamed &&
1595 (added + deleted == 0)) {
1596 total_files--;
1597 continue;
1598 }
1599 /*
1600 * "scale" the filename
1601 */
1602 len = name_width;
1603 name_len = strlen(name);
1604 if (name_width < name_len) {
1605 char *slash;
1606 prefix = "...";
1607 len -= 3;
1608 name += name_len - len;
1609 slash = strchr(name, '/');
1610 if (slash)
1611 name = slash;
1612 }
1613
1614 if (data->files[i]->is_binary) {
1615 fprintf(options->file, "%s", line_prefix);
1616 show_name(options->file, prefix, name, len);
1617 fprintf(options->file, " %*s ", number_width, "Bin");
1618 fprintf(options->file, "%s%"PRIuMAX"%s",
1619 del_c, deleted, reset);
1620 fprintf(options->file, " -> ");
1621 fprintf(options->file, "%s%"PRIuMAX"%s",
1622 add_c, added, reset);
1623 fprintf(options->file, " bytes");
1624 fprintf(options->file, "\n");
1625 continue;
1626 }
1627 else if (data->files[i]->is_unmerged) {
1628 fprintf(options->file, "%s", line_prefix);
1629 show_name(options->file, prefix, name, len);
1630 fprintf(options->file, " Unmerged\n");
1631 continue;
1632 }
1633
1634 /*
1635 * scale the add/delete
1636 */
1637 add = added;
1638 del = deleted;
1639 adds += add;
1640 dels += del;
1641
1642 if (graph_width <= max_change) {
1643 int total = add + del;
1644
1645 total = scale_linear(add + del, graph_width, max_change);
1646 if (total < 2 && add && del)
1647 /* width >= 2 due to the sanity check */
1648 total = 2;
1649 if (add < del) {
1650 add = scale_linear(add, graph_width, max_change);
1651 del = total - add;
1652 } else {
1653 del = scale_linear(del, graph_width, max_change);
1654 add = total - del;
1655 }
1656 }
1657 fprintf(options->file, "%s", line_prefix);
1658 show_name(options->file, prefix, name, len);
1659 fprintf(options->file, " %*"PRIuMAX"%s",
1660 number_width, added + deleted,
1661 added + deleted ? " " : "");
1662 show_graph(options->file, '+', add, add_c, reset);
1663 show_graph(options->file, '-', del, del_c, reset);
1664 fprintf(options->file, "\n");
1665 }
1666 for (i = count; i < data->nr; i++) {
1667 uintmax_t added = data->files[i]->added;
1668 uintmax_t deleted = data->files[i]->deleted;
1669 if (!data->files[i]->is_renamed &&
1670 (added + deleted == 0)) {
1671 total_files--;
1672 continue;
1673 }
1674 adds += added;
1675 dels += deleted;
1676 if (!extra_shown)
1677 fprintf(options->file, "%s ...\n", line_prefix);
1678 extra_shown = 1;
1679 }
1680 fprintf(options->file, "%s", line_prefix);
1681 print_stat_summary(options->file, total_files, adds, dels);
1682}
1683
1684static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1685{
1686 int i, adds = 0, dels = 0, total_files = data->nr;
1687
1688 if (data->nr == 0)
1689 return;
1690
1691 for (i = 0; i < data->nr; i++) {
1692 if (!data->files[i]->is_binary &&
1693 !data->files[i]->is_unmerged) {
1694 int added = data->files[i]->added;
1695 int deleted= data->files[i]->deleted;
1696 if (!data->files[i]->is_renamed &&
1697 (added + deleted == 0)) {
1698 total_files--;
1699 } else {
1700 adds += added;
1701 dels += deleted;
1702 }
1703 }
1704 }
1705 if (options->output_prefix) {
1706 struct strbuf *msg = NULL;
1707 msg = options->output_prefix(options,
1708 options->output_prefix_data);
1709 fprintf(options->file, "%s", msg->buf);
1710 }
1711 print_stat_summary(options->file, total_files, adds, dels);
1712}
1713
1714static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1715{
1716 int i;
1717
1718 if (data->nr == 0)
1719 return;
1720
1721 for (i = 0; i < data->nr; i++) {
1722 struct diffstat_file *file = data->files[i];
1723
1724 if (options->output_prefix) {
1725 struct strbuf *msg = NULL;
1726 msg = options->output_prefix(options,
1727 options->output_prefix_data);
1728 fprintf(options->file, "%s", msg->buf);
1729 }
1730
1731 if (file->is_binary)
1732 fprintf(options->file, "-\t-\t");
1733 else
1734 fprintf(options->file,
1735 "%"PRIuMAX"\t%"PRIuMAX"\t",
1736 file->added, file->deleted);
1737 if (options->line_termination) {
1738 fill_print_name(file);
1739 if (!file->is_renamed)
1740 write_name_quoted(file->name, options->file,
1741 options->line_termination);
1742 else {
1743 fputs(file->print_name, options->file);
1744 putc(options->line_termination, options->file);
1745 }
1746 } else {
1747 if (file->is_renamed) {
1748 putc('\0', options->file);
1749 write_name_quoted(file->from_name, options->file, '\0');
1750 }
1751 write_name_quoted(file->name, options->file, '\0');
1752 }
1753 }
1754}
1755
1756struct dirstat_file {
1757 const char *name;
1758 unsigned long changed;
1759};
1760
1761struct dirstat_dir {
1762 struct dirstat_file *files;
1763 int alloc, nr, permille, cumulative;
1764};
1765
1766static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1767 unsigned long changed, const char *base, int baselen)
1768{
1769 unsigned long this_dir = 0;
1770 unsigned int sources = 0;
1771 const char *line_prefix = "";
1772 struct strbuf *msg = NULL;
1773
1774 if (opt->output_prefix) {
1775 msg = opt->output_prefix(opt, opt->output_prefix_data);
1776 line_prefix = msg->buf;
1777 }
1778
1779 while (dir->nr) {
1780 struct dirstat_file *f = dir->files;
1781 int namelen = strlen(f->name);
1782 unsigned long this;
1783 char *slash;
1784
1785 if (namelen < baselen)
1786 break;
1787 if (memcmp(f->name, base, baselen))
1788 break;
1789 slash = strchr(f->name + baselen, '/');
1790 if (slash) {
1791 int newbaselen = slash + 1 - f->name;
1792 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1793 sources++;
1794 } else {
1795 this = f->changed;
1796 dir->files++;
1797 dir->nr--;
1798 sources += 2;
1799 }
1800 this_dir += this;
1801 }
1802
1803 /*
1804 * We don't report dirstat's for
1805 * - the top level
1806 * - or cases where everything came from a single directory
1807 * under this directory (sources == 1).
1808 */
1809 if (baselen && sources != 1) {
1810 if (this_dir) {
1811 int permille = this_dir * 1000 / changed;
1812 if (permille >= dir->permille) {
1813 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1814 permille / 10, permille % 10, baselen, base);
1815 if (!dir->cumulative)
1816 return 0;
1817 }
1818 }
1819 }
1820 return this_dir;
1821}
1822
1823static int dirstat_compare(const void *_a, const void *_b)
1824{
1825 const struct dirstat_file *a = _a;
1826 const struct dirstat_file *b = _b;
1827 return strcmp(a->name, b->name);
1828}
1829
1830static void show_dirstat(struct diff_options *options)
1831{
1832 int i;
1833 unsigned long changed;
1834 struct dirstat_dir dir;
1835 struct diff_queue_struct *q = &diff_queued_diff;
1836
1837 dir.files = NULL;
1838 dir.alloc = 0;
1839 dir.nr = 0;
1840 dir.permille = options->dirstat_permille;
1841 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1842
1843 changed = 0;
1844 for (i = 0; i < q->nr; i++) {
1845 struct diff_filepair *p = q->queue[i];
1846 const char *name;
1847 unsigned long copied, added, damage;
1848 int content_changed;
1849
1850 name = p->two->path ? p->two->path : p->one->path;
1851
1852 if (p->one->sha1_valid && p->two->sha1_valid)
1853 content_changed = hashcmp(p->one->sha1, p->two->sha1);
1854 else
1855 content_changed = 1;
1856
1857 if (!content_changed) {
1858 /*
1859 * The SHA1 has not changed, so pre-/post-content is
1860 * identical. We can therefore skip looking at the
1861 * file contents altogether.
1862 */
1863 damage = 0;
1864 goto found_damage;
1865 }
1866
1867 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1868 /*
1869 * In --dirstat-by-file mode, we don't really need to
1870 * look at the actual file contents at all.
1871 * The fact that the SHA1 changed is enough for us to
1872 * add this file to the list of results
1873 * (with each file contributing equal damage).
1874 */
1875 damage = 1;
1876 goto found_damage;
1877 }
1878
1879 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1880 diff_populate_filespec(p->one, 0);
1881 diff_populate_filespec(p->two, 0);
1882 diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1883 &copied, &added);
1884 diff_free_filespec_data(p->one);
1885 diff_free_filespec_data(p->two);
1886 } else if (DIFF_FILE_VALID(p->one)) {
1887 diff_populate_filespec(p->one, 1);
1888 copied = added = 0;
1889 diff_free_filespec_data(p->one);
1890 } else if (DIFF_FILE_VALID(p->two)) {
1891 diff_populate_filespec(p->two, 1);
1892 copied = 0;
1893 added = p->two->size;
1894 diff_free_filespec_data(p->two);
1895 } else
1896 continue;
1897
1898 /*
1899 * Original minus copied is the removed material,
1900 * added is the new material. They are both damages
1901 * made to the preimage.
1902 * If the resulting damage is zero, we know that
1903 * diffcore_count_changes() considers the two entries to
1904 * be identical, but since content_changed is true, we
1905 * know that there must have been _some_ kind of change,
1906 * so we force all entries to have damage > 0.
1907 */
1908 damage = (p->one->size - copied) + added;
1909 if (!damage)
1910 damage = 1;
1911
1912found_damage:
1913 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1914 dir.files[dir.nr].name = name;
1915 dir.files[dir.nr].changed = damage;
1916 changed += damage;
1917 dir.nr++;
1918 }
1919
1920 /* This can happen even with many files, if everything was renames */
1921 if (!changed)
1922 return;
1923
1924 /* Show all directories with more than x% of the changes */
1925 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1926 gather_dirstat(options, &dir, changed, "", 0);
1927}
1928
1929static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
1930{
1931 int i;
1932 unsigned long changed;
1933 struct dirstat_dir dir;
1934
1935 if (data->nr == 0)
1936 return;
1937
1938 dir.files = NULL;
1939 dir.alloc = 0;
1940 dir.nr = 0;
1941 dir.permille = options->dirstat_permille;
1942 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1943
1944 changed = 0;
1945 for (i = 0; i < data->nr; i++) {
1946 struct diffstat_file *file = data->files[i];
1947 unsigned long damage = file->added + file->deleted;
1948 if (file->is_binary)
1949 /*
1950 * binary files counts bytes, not lines. Must find some
1951 * way to normalize binary bytes vs. textual lines.
1952 * The following heuristic assumes that there are 64
1953 * bytes per "line".
1954 * This is stupid and ugly, but very cheap...
1955 */
1956 damage = (damage + 63) / 64;
1957 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1958 dir.files[dir.nr].name = file->name;
1959 dir.files[dir.nr].changed = damage;
1960 changed += damage;
1961 dir.nr++;
1962 }
1963
1964 /* This can happen even with many files, if everything was renames */
1965 if (!changed)
1966 return;
1967
1968 /* Show all directories with more than x% of the changes */
1969 qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1970 gather_dirstat(options, &dir, changed, "", 0);
1971}
1972
1973static void free_diffstat_info(struct diffstat_t *diffstat)
1974{
1975 int i;
1976 for (i = 0; i < diffstat->nr; i++) {
1977 struct diffstat_file *f = diffstat->files[i];
1978 if (f->name != f->print_name)
1979 free(f->print_name);
1980 free(f->name);
1981 free(f->from_name);
1982 free(f);
1983 }
1984 free(diffstat->files);
1985}
1986
1987struct checkdiff_t {
1988 const char *filename;
1989 int lineno;
1990 int conflict_marker_size;
1991 struct diff_options *o;
1992 unsigned ws_rule;
1993 unsigned status;
1994};
1995
1996static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
1997{
1998 char firstchar;
1999 int cnt;
2000
2001 if (len < marker_size + 1)
2002 return 0;
2003 firstchar = line[0];
2004 switch (firstchar) {
2005 case '=': case '>': case '<': case '|':
2006 break;
2007 default:
2008 return 0;
2009 }
2010 for (cnt = 1; cnt < marker_size; cnt++)
2011 if (line[cnt] != firstchar)
2012 return 0;
2013 /* line[1] thru line[marker_size-1] are same as firstchar */
2014 if (len < marker_size + 1 || !isspace(line[marker_size]))
2015 return 0;
2016 return 1;
2017}
2018
2019static void checkdiff_consume(void *priv, char *line, unsigned long len)
2020{
2021 struct checkdiff_t *data = priv;
2022 int marker_size = data->conflict_marker_size;
2023 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2024 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2025 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2026 char *err;
2027 char *line_prefix = "";
2028 struct strbuf *msgbuf;
2029
2030 assert(data->o);
2031 if (data->o->output_prefix) {
2032 msgbuf = data->o->output_prefix(data->o,
2033 data->o->output_prefix_data);
2034 line_prefix = msgbuf->buf;
2035 }
2036
2037 if (line[0] == '+') {
2038 unsigned bad;
2039 data->lineno++;
2040 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2041 data->status |= 1;
2042 fprintf(data->o->file,
2043 "%s%s:%d: leftover conflict marker\n",
2044 line_prefix, data->filename, data->lineno);
2045 }
2046 bad = ws_check(line + 1, len - 1, data->ws_rule);
2047 if (!bad)
2048 return;
2049 data->status |= bad;
2050 err = whitespace_error_string(bad);
2051 fprintf(data->o->file, "%s%s:%d: %s.\n",
2052 line_prefix, data->filename, data->lineno, err);
2053 free(err);
2054 emit_line(data->o, set, reset, line, 1);
2055 ws_check_emit(line + 1, len - 1, data->ws_rule,
2056 data->o->file, set, reset, ws);
2057 } else if (line[0] == ' ') {
2058 data->lineno++;
2059 } else if (line[0] == '@') {
2060 char *plus = strchr(line, '+');
2061 if (plus)
2062 data->lineno = strtol(plus, NULL, 10) - 1;
2063 else
2064 die("invalid diff");
2065 }
2066}
2067
2068static unsigned char *deflate_it(char *data,
2069 unsigned long size,
2070 unsigned long *result_size)
2071{
2072 int bound;
2073 unsigned char *deflated;
2074 git_zstream stream;
2075
2076 memset(&stream, 0, sizeof(stream));
2077 git_deflate_init(&stream, zlib_compression_level);
2078 bound = git_deflate_bound(&stream, size);
2079 deflated = xmalloc(bound);
2080 stream.next_out = deflated;
2081 stream.avail_out = bound;
2082
2083 stream.next_in = (unsigned char *)data;
2084 stream.avail_in = size;
2085 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2086 ; /* nothing */
2087 git_deflate_end(&stream);
2088 *result_size = stream.total_out;
2089 return deflated;
2090}
2091
2092static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
2093{
2094 void *cp;
2095 void *delta;
2096 void *deflated;
2097 void *data;
2098 unsigned long orig_size;
2099 unsigned long delta_size;
2100 unsigned long deflate_size;
2101 unsigned long data_size;
2102
2103 /* We could do deflated delta, or we could do just deflated two,
2104 * whichever is smaller.
2105 */
2106 delta = NULL;
2107 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2108 if (one->size && two->size) {
2109 delta = diff_delta(one->ptr, one->size,
2110 two->ptr, two->size,
2111 &delta_size, deflate_size);
2112 if (delta) {
2113 void *to_free = delta;
2114 orig_size = delta_size;
2115 delta = deflate_it(delta, delta_size, &delta_size);
2116 free(to_free);
2117 }
2118 }
2119
2120 if (delta && delta_size < deflate_size) {
2121 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2122 free(deflated);
2123 data = delta;
2124 data_size = delta_size;
2125 }
2126 else {
2127 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2128 free(delta);
2129 data = deflated;
2130 data_size = deflate_size;
2131 }
2132
2133 /* emit data encoded in base85 */
2134 cp = data;
2135 while (data_size) {
2136 int bytes = (52 < data_size) ? 52 : data_size;
2137 char line[70];
2138 data_size -= bytes;
2139 if (bytes <= 26)
2140 line[0] = bytes + 'A' - 1;
2141 else
2142 line[0] = bytes - 26 + 'a' - 1;
2143 encode_85(line + 1, cp, bytes);
2144 cp = (char *) cp + bytes;
2145 fprintf(file, "%s", prefix);
2146 fputs(line, file);
2147 fputc('\n', file);
2148 }
2149 fprintf(file, "%s\n", prefix);
2150 free(data);
2151}
2152
2153static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
2154{
2155 fprintf(file, "%sGIT binary patch\n", prefix);
2156 emit_binary_diff_body(file, one, two, prefix);
2157 emit_binary_diff_body(file, two, one, prefix);
2158}
2159
2160int diff_filespec_is_binary(struct diff_filespec *one)
2161{
2162 if (one->is_binary == -1) {
2163 diff_filespec_load_driver(one);
2164 if (one->driver->binary != -1)
2165 one->is_binary = one->driver->binary;
2166 else {
2167 if (!one->data && DIFF_FILE_VALID(one))
2168 diff_populate_filespec(one, 0);
2169 if (one->data)
2170 one->is_binary = buffer_is_binary(one->data,
2171 one->size);
2172 if (one->is_binary == -1)
2173 one->is_binary = 0;
2174 }
2175 }
2176 return one->is_binary;
2177}
2178
2179static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2180{
2181 diff_filespec_load_driver(one);
2182 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2183}
2184
2185void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2186{
2187 if (!options->a_prefix)
2188 options->a_prefix = a;
2189 if (!options->b_prefix)
2190 options->b_prefix = b;
2191}
2192
2193struct userdiff_driver *get_textconv(struct diff_filespec *one)
2194{
2195 if (!DIFF_FILE_VALID(one))
2196 return NULL;
2197
2198 diff_filespec_load_driver(one);
2199 return userdiff_get_textconv(one->driver);
2200}
2201
2202static void builtin_diff(const char *name_a,
2203 const char *name_b,
2204 struct diff_filespec *one,
2205 struct diff_filespec *two,
2206 const char *xfrm_msg,
2207 int must_show_header,
2208 struct diff_options *o,
2209 int complete_rewrite)
2210{
2211 mmfile_t mf1, mf2;
2212 const char *lbl[2];
2213 char *a_one, *b_two;
2214 const char *set = diff_get_color_opt(o, DIFF_METAINFO);
2215 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2216 const char *a_prefix, *b_prefix;
2217 struct userdiff_driver *textconv_one = NULL;
2218 struct userdiff_driver *textconv_two = NULL;
2219 struct strbuf header = STRBUF_INIT;
2220 struct strbuf *msgbuf;
2221 char *line_prefix = "";
2222
2223 if (o->output_prefix) {
2224 msgbuf = o->output_prefix(o, o->output_prefix_data);
2225 line_prefix = msgbuf->buf;
2226 }
2227
2228 if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2229 (!one->mode || S_ISGITLINK(one->mode)) &&
2230 (!two->mode || S_ISGITLINK(two->mode))) {
2231 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2232 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2233 show_submodule_summary(o->file, one ? one->path : two->path,
2234 one->sha1, two->sha1, two->dirty_submodule,
2235 del, add, reset);
2236 return;
2237 }
2238
2239 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2240 textconv_one = get_textconv(one);
2241 textconv_two = get_textconv(two);
2242 }
2243
2244 diff_set_mnemonic_prefix(o, "a/", "b/");
2245 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2246 a_prefix = o->b_prefix;
2247 b_prefix = o->a_prefix;
2248 } else {
2249 a_prefix = o->a_prefix;
2250 b_prefix = o->b_prefix;
2251 }
2252
2253 /* Never use a non-valid filename anywhere if at all possible */
2254 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2255 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2256
2257 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2258 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2259 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2260 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2261 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, set, a_one, b_two, reset);
2262 if (lbl[0][0] == '/') {
2263 /* /dev/null */
2264 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, set, two->mode, reset);
2265 if (xfrm_msg)
2266 strbuf_addstr(&header, xfrm_msg);
2267 must_show_header = 1;
2268 }
2269 else if (lbl[1][0] == '/') {
2270 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, set, one->mode, reset);
2271 if (xfrm_msg)
2272 strbuf_addstr(&header, xfrm_msg);
2273 must_show_header = 1;
2274 }
2275 else {
2276 if (one->mode != two->mode) {
2277 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, set, one->mode, reset);
2278 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, set, two->mode, reset);
2279 must_show_header = 1;
2280 }
2281 if (xfrm_msg)
2282 strbuf_addstr(&header, xfrm_msg);
2283
2284 /*
2285 * we do not run diff between different kind
2286 * of objects.
2287 */
2288 if ((one->mode ^ two->mode) & S_IFMT)
2289 goto free_ab_and_return;
2290 if (complete_rewrite &&
2291 (textconv_one || !diff_filespec_is_binary(one)) &&
2292 (textconv_two || !diff_filespec_is_binary(two))) {
2293 fprintf(o->file, "%s", header.buf);
2294 strbuf_reset(&header);
2295 emit_rewrite_diff(name_a, name_b, one, two,
2296 textconv_one, textconv_two, o);
2297 o->found_changes = 1;
2298 goto free_ab_and_return;
2299 }
2300 }
2301
2302 if (o->irreversible_delete && lbl[1][0] == '/') {
2303 fprintf(o->file, "%s", header.buf);
2304 strbuf_reset(&header);
2305 goto free_ab_and_return;
2306 } else if (!DIFF_OPT_TST(o, TEXT) &&
2307 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2308 (!textconv_two && diff_filespec_is_binary(two)) )) {
2309 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2310 die("unable to read files to diff");
2311 /* Quite common confusing case */
2312 if (mf1.size == mf2.size &&
2313 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2314 if (must_show_header)
2315 fprintf(o->file, "%s", header.buf);
2316 goto free_ab_and_return;
2317 }
2318 fprintf(o->file, "%s", header.buf);
2319 strbuf_reset(&header);
2320 if (DIFF_OPT_TST(o, BINARY))
2321 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2322 else
2323 fprintf(o->file, "%sBinary files %s and %s differ\n",
2324 line_prefix, lbl[0], lbl[1]);
2325 o->found_changes = 1;
2326 } else {
2327 /* Crazy xdl interfaces.. */
2328 const char *diffopts = getenv("GIT_DIFF_OPTS");
2329 xpparam_t xpp;
2330 xdemitconf_t xecfg;
2331 struct emit_callback ecbdata;
2332 const struct userdiff_funcname *pe;
2333
2334 if (must_show_header) {
2335 fprintf(o->file, "%s", header.buf);
2336 strbuf_reset(&header);
2337 }
2338
2339 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2340 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2341
2342 pe = diff_funcname_pattern(one);
2343 if (!pe)
2344 pe = diff_funcname_pattern(two);
2345
2346 memset(&xpp, 0, sizeof(xpp));
2347 memset(&xecfg, 0, sizeof(xecfg));
2348 memset(&ecbdata, 0, sizeof(ecbdata));
2349 ecbdata.label_path = lbl;
2350 ecbdata.color_diff = want_color(o->use_color);
2351 ecbdata.found_changesp = &o->found_changes;
2352 ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
2353 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2354 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2355 ecbdata.opt = o;
2356 ecbdata.header = header.len ? &header : NULL;
2357 xpp.flags = o->xdl_opts;
2358 xecfg.ctxlen = o->context;
2359 xecfg.interhunkctxlen = o->interhunkcontext;
2360 xecfg.flags = XDL_EMIT_FUNCNAMES;
2361 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2362 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2363 if (pe)
2364 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2365 if (!diffopts)
2366 ;
2367 else if (!prefixcmp(diffopts, "--unified="))
2368 xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
2369 else if (!prefixcmp(diffopts, "-u"))
2370 xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
2371 if (o->word_diff)
2372 init_diff_words_data(&ecbdata, o, one, two);
2373 xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2374 &xpp, &xecfg);
2375 if (o->word_diff)
2376 free_diff_words_data(&ecbdata);
2377 if (textconv_one)
2378 free(mf1.ptr);
2379 if (textconv_two)
2380 free(mf2.ptr);
2381 xdiff_clear_find_func(&xecfg);
2382 }
2383
2384 free_ab_and_return:
2385 strbuf_release(&header);
2386 diff_free_filespec_data(one);
2387 diff_free_filespec_data(two);
2388 free(a_one);
2389 free(b_two);
2390 return;
2391}
2392
2393static void builtin_diffstat(const char *name_a, const char *name_b,
2394 struct diff_filespec *one,
2395 struct diff_filespec *two,
2396 struct diffstat_t *diffstat,
2397 struct diff_options *o,
2398 int complete_rewrite)
2399{
2400 mmfile_t mf1, mf2;
2401 struct diffstat_file *data;
2402
2403 data = diffstat_add(diffstat, name_a, name_b);
2404
2405 if (!one || !two) {
2406 data->is_unmerged = 1;
2407 return;
2408 }
2409
2410 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2411 data->is_binary = 1;
2412 data->added = diff_filespec_size(two);
2413 data->deleted = diff_filespec_size(one);
2414 }
2415
2416 else if (complete_rewrite) {
2417 diff_populate_filespec(one, 0);
2418 diff_populate_filespec(two, 0);
2419 data->deleted = count_lines(one->data, one->size);
2420 data->added = count_lines(two->data, two->size);
2421 }
2422
2423 else {
2424 /* Crazy xdl interfaces.. */
2425 xpparam_t xpp;
2426 xdemitconf_t xecfg;
2427
2428 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2429 die("unable to read files to diff");
2430
2431 memset(&xpp, 0, sizeof(xpp));
2432 memset(&xecfg, 0, sizeof(xecfg));
2433 xpp.flags = o->xdl_opts;
2434 xecfg.ctxlen = o->context;
2435 xecfg.interhunkctxlen = o->interhunkcontext;
2436 xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2437 &xpp, &xecfg);
2438 }
2439
2440 diff_free_filespec_data(one);
2441 diff_free_filespec_data(two);
2442}
2443
2444static void builtin_checkdiff(const char *name_a, const char *name_b,
2445 const char *attr_path,
2446 struct diff_filespec *one,
2447 struct diff_filespec *two,
2448 struct diff_options *o)
2449{
2450 mmfile_t mf1, mf2;
2451 struct checkdiff_t data;
2452
2453 if (!two)
2454 return;
2455
2456 memset(&data, 0, sizeof(data));
2457 data.filename = name_b ? name_b : name_a;
2458 data.lineno = 0;
2459 data.o = o;
2460 data.ws_rule = whitespace_rule(attr_path);
2461 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2462
2463 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2464 die("unable to read files to diff");
2465
2466 /*
2467 * All the other codepaths check both sides, but not checking
2468 * the "old" side here is deliberate. We are checking the newly
2469 * introduced changes, and as long as the "new" side is text, we
2470 * can and should check what it introduces.
2471 */
2472 if (diff_filespec_is_binary(two))
2473 goto free_and_return;
2474 else {
2475 /* Crazy xdl interfaces.. */
2476 xpparam_t xpp;
2477 xdemitconf_t xecfg;
2478
2479 memset(&xpp, 0, sizeof(xpp));
2480 memset(&xecfg, 0, sizeof(xecfg));
2481 xecfg.ctxlen = 1; /* at least one context line */
2482 xpp.flags = 0;
2483 xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2484 &xpp, &xecfg);
2485
2486 if (data.ws_rule & WS_BLANK_AT_EOF) {
2487 struct emit_callback ecbdata;
2488 int blank_at_eof;
2489
2490 ecbdata.ws_rule = data.ws_rule;
2491 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2492 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2493
2494 if (blank_at_eof) {
2495 static char *err;
2496 if (!err)
2497 err = whitespace_error_string(WS_BLANK_AT_EOF);
2498 fprintf(o->file, "%s:%d: %s.\n",
2499 data.filename, blank_at_eof, err);
2500 data.status = 1; /* report errors */
2501 }
2502 }
2503 }
2504 free_and_return:
2505 diff_free_filespec_data(one);
2506 diff_free_filespec_data(two);
2507 if (data.status)
2508 DIFF_OPT_SET(o, CHECK_FAILED);
2509}
2510
2511struct diff_filespec *alloc_filespec(const char *path)
2512{
2513 int namelen = strlen(path);
2514 struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
2515
2516 memset(spec, 0, sizeof(*spec));
2517 spec->path = (char *)(spec + 1);
2518 memcpy(spec->path, path, namelen+1);
2519 spec->count = 1;
2520 spec->is_binary = -1;
2521 return spec;
2522}
2523
2524void free_filespec(struct diff_filespec *spec)
2525{
2526 if (!--spec->count) {
2527 diff_free_filespec_data(spec);
2528 free(spec);
2529 }
2530}
2531
2532void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2533 unsigned short mode)
2534{
2535 if (mode) {
2536 spec->mode = canon_mode(mode);
2537 hashcpy(spec->sha1, sha1);
2538 spec->sha1_valid = !is_null_sha1(sha1);
2539 }
2540}
2541
2542/*
2543 * Given a name and sha1 pair, if the index tells us the file in
2544 * the work tree has that object contents, return true, so that
2545 * prepare_temp_file() does not have to inflate and extract.
2546 */
2547static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2548{
2549 struct cache_entry *ce;
2550 struct stat st;
2551 int pos, len;
2552
2553 /*
2554 * We do not read the cache ourselves here, because the
2555 * benchmark with my previous version that always reads cache
2556 * shows that it makes things worse for diff-tree comparing
2557 * two linux-2.6 kernel trees in an already checked out work
2558 * tree. This is because most diff-tree comparisons deal with
2559 * only a small number of files, while reading the cache is
2560 * expensive for a large project, and its cost outweighs the
2561 * savings we get by not inflating the object to a temporary
2562 * file. Practically, this code only helps when we are used
2563 * by diff-cache --cached, which does read the cache before
2564 * calling us.
2565 */
2566 if (!active_cache)
2567 return 0;
2568
2569 /* We want to avoid the working directory if our caller
2570 * doesn't need the data in a normal file, this system
2571 * is rather slow with its stat/open/mmap/close syscalls,
2572 * and the object is contained in a pack file. The pack
2573 * is probably already open and will be faster to obtain
2574 * the data through than the working directory. Loose
2575 * objects however would tend to be slower as they need
2576 * to be individually opened and inflated.
2577 */
2578 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2579 return 0;
2580
2581 len = strlen(name);
2582 pos = cache_name_pos(name, len);
2583 if (pos < 0)
2584 return 0;
2585 ce = active_cache[pos];
2586
2587 /*
2588 * This is not the sha1 we are looking for, or
2589 * unreusable because it is not a regular file.
2590 */
2591 if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2592 return 0;
2593
2594 /*
2595 * If ce is marked as "assume unchanged", there is no
2596 * guarantee that work tree matches what we are looking for.
2597 */
2598 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2599 return 0;
2600
2601 /*
2602 * If ce matches the file in the work tree, we can reuse it.
2603 */
2604 if (ce_uptodate(ce) ||
2605 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2606 return 1;
2607
2608 return 0;
2609}
2610
2611static int populate_from_stdin(struct diff_filespec *s)
2612{
2613 struct strbuf buf = STRBUF_INIT;
2614 size_t size = 0;
2615
2616 if (strbuf_read(&buf, 0, 0) < 0)
2617 return error("error while reading from stdin %s",
2618 strerror(errno));
2619
2620 s->should_munmap = 0;
2621 s->data = strbuf_detach(&buf, &size);
2622 s->size = size;
2623 s->should_free = 1;
2624 return 0;
2625}
2626
2627static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2628{
2629 int len;
2630 char *data = xmalloc(100), *dirty = "";
2631
2632 /* Are we looking at the work tree? */
2633 if (s->dirty_submodule)
2634 dirty = "-dirty";
2635
2636 len = snprintf(data, 100,
2637 "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2638 s->data = data;
2639 s->size = len;
2640 s->should_free = 1;
2641 if (size_only) {
2642 s->data = NULL;
2643 free(data);
2644 }
2645 return 0;
2646}
2647
2648/*
2649 * While doing rename detection and pickaxe operation, we may need to
2650 * grab the data for the blob (or file) for our own in-core comparison.
2651 * diff_filespec has data and size fields for this purpose.
2652 */
2653int diff_populate_filespec(struct diff_filespec *s, int size_only)
2654{
2655 int err = 0;
2656 if (!DIFF_FILE_VALID(s))
2657 die("internal error: asking to populate invalid file.");
2658 if (S_ISDIR(s->mode))
2659 return -1;
2660
2661 if (s->data)
2662 return 0;
2663
2664 if (size_only && 0 < s->size)
2665 return 0;
2666
2667 if (S_ISGITLINK(s->mode))
2668 return diff_populate_gitlink(s, size_only);
2669
2670 if (!s->sha1_valid ||
2671 reuse_worktree_file(s->path, s->sha1, 0)) {
2672 struct strbuf buf = STRBUF_INIT;
2673 struct stat st;
2674 int fd;
2675
2676 if (!strcmp(s->path, "-"))
2677 return populate_from_stdin(s);
2678
2679 if (lstat(s->path, &st) < 0) {
2680 if (errno == ENOENT) {
2681 err_empty:
2682 err = -1;
2683 empty:
2684 s->data = (char *)"";
2685 s->size = 0;
2686 return err;
2687 }
2688 }
2689 s->size = xsize_t(st.st_size);
2690 if (!s->size)
2691 goto empty;
2692 if (S_ISLNK(st.st_mode)) {
2693 struct strbuf sb = STRBUF_INIT;
2694
2695 if (strbuf_readlink(&sb, s->path, s->size))
2696 goto err_empty;
2697 s->size = sb.len;
2698 s->data = strbuf_detach(&sb, NULL);
2699 s->should_free = 1;
2700 return 0;
2701 }
2702 if (size_only)
2703 return 0;
2704 fd = open(s->path, O_RDONLY);
2705 if (fd < 0)
2706 goto err_empty;
2707 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2708 close(fd);
2709 s->should_munmap = 1;
2710
2711 /*
2712 * Convert from working tree format to canonical git format
2713 */
2714 if (convert_to_git(s->path, s->data, s->size, &buf, safe_crlf)) {
2715 size_t size = 0;
2716 munmap(s->data, s->size);
2717 s->should_munmap = 0;
2718 s->data = strbuf_detach(&buf, &size);
2719 s->size = size;
2720 s->should_free = 1;
2721 }
2722 }
2723 else {
2724 enum object_type type;
2725 if (size_only) {
2726 type = sha1_object_info(s->sha1, &s->size);
2727 if (type < 0)
2728 die("unable to read %s", sha1_to_hex(s->sha1));
2729 } else {
2730 s->data = read_sha1_file(s->sha1, &type, &s->size);
2731 if (!s->data)
2732 die("unable to read %s", sha1_to_hex(s->sha1));
2733 s->should_free = 1;
2734 }
2735 }
2736 return 0;
2737}
2738
2739void diff_free_filespec_blob(struct diff_filespec *s)
2740{
2741 if (s->should_free)
2742 free(s->data);
2743 else if (s->should_munmap)
2744 munmap(s->data, s->size);
2745
2746 if (s->should_free || s->should_munmap) {
2747 s->should_free = s->should_munmap = 0;
2748 s->data = NULL;
2749 }
2750}
2751
2752void diff_free_filespec_data(struct diff_filespec *s)
2753{
2754 diff_free_filespec_blob(s);
2755 free(s->cnt_data);
2756 s->cnt_data = NULL;
2757}
2758
2759static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2760 void *blob,
2761 unsigned long size,
2762 const unsigned char *sha1,
2763 int mode)
2764{
2765 int fd;
2766 struct strbuf buf = STRBUF_INIT;
2767 struct strbuf template = STRBUF_INIT;
2768 char *path_dup = xstrdup(path);
2769 const char *base = basename(path_dup);
2770
2771 /* Generate "XXXXXX_basename.ext" */
2772 strbuf_addstr(&template, "XXXXXX_");
2773 strbuf_addstr(&template, base);
2774
2775 fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
2776 strlen(base) + 1);
2777 if (fd < 0)
2778 die_errno("unable to create temp-file");
2779 if (convert_to_working_tree(path,
2780 (const char *)blob, (size_t)size, &buf)) {
2781 blob = buf.buf;
2782 size = buf.len;
2783 }
2784 if (write_in_full(fd, blob, size) != size)
2785 die_errno("unable to write temp-file");
2786 close(fd);
2787 temp->name = temp->tmp_path;
2788 strcpy(temp->hex, sha1_to_hex(sha1));
2789 temp->hex[40] = 0;
2790 sprintf(temp->mode, "%06o", mode);
2791 strbuf_release(&buf);
2792 strbuf_release(&template);
2793 free(path_dup);
2794}
2795
2796static struct diff_tempfile *prepare_temp_file(const char *name,
2797 struct diff_filespec *one)
2798{
2799 struct diff_tempfile *temp = claim_diff_tempfile();
2800
2801 if (!DIFF_FILE_VALID(one)) {
2802 not_a_valid_file:
2803 /* A '-' entry produces this for file-2, and
2804 * a '+' entry produces this for file-1.
2805 */
2806 temp->name = "/dev/null";
2807 strcpy(temp->hex, ".");
2808 strcpy(temp->mode, ".");
2809 return temp;
2810 }
2811
2812 if (!remove_tempfile_installed) {
2813 atexit(remove_tempfile);
2814 sigchain_push_common(remove_tempfile_on_signal);
2815 remove_tempfile_installed = 1;
2816 }
2817
2818 if (!one->sha1_valid ||
2819 reuse_worktree_file(name, one->sha1, 1)) {
2820 struct stat st;
2821 if (lstat(name, &st) < 0) {
2822 if (errno == ENOENT)
2823 goto not_a_valid_file;
2824 die_errno("stat(%s)", name);
2825 }
2826 if (S_ISLNK(st.st_mode)) {
2827 struct strbuf sb = STRBUF_INIT;
2828 if (strbuf_readlink(&sb, name, st.st_size) < 0)
2829 die_errno("readlink(%s)", name);
2830 prep_temp_blob(name, temp, sb.buf, sb.len,
2831 (one->sha1_valid ?
2832 one->sha1 : null_sha1),
2833 (one->sha1_valid ?
2834 one->mode : S_IFLNK));
2835 strbuf_release(&sb);
2836 }
2837 else {
2838 /* we can borrow from the file in the work tree */
2839 temp->name = name;
2840 if (!one->sha1_valid)
2841 strcpy(temp->hex, sha1_to_hex(null_sha1));
2842 else
2843 strcpy(temp->hex, sha1_to_hex(one->sha1));
2844 /* Even though we may sometimes borrow the
2845 * contents from the work tree, we always want
2846 * one->mode. mode is trustworthy even when
2847 * !(one->sha1_valid), as long as
2848 * DIFF_FILE_VALID(one).
2849 */
2850 sprintf(temp->mode, "%06o", one->mode);
2851 }
2852 return temp;
2853 }
2854 else {
2855 if (diff_populate_filespec(one, 0))
2856 die("cannot read data blob for %s", one->path);
2857 prep_temp_blob(name, temp, one->data, one->size,
2858 one->sha1, one->mode);
2859 }
2860 return temp;
2861}
2862
2863/* An external diff command takes:
2864 *
2865 * diff-cmd name infile1 infile1-sha1 infile1-mode \
2866 * infile2 infile2-sha1 infile2-mode [ rename-to ]
2867 *
2868 */
2869static void run_external_diff(const char *pgm,
2870 const char *name,
2871 const char *other,
2872 struct diff_filespec *one,
2873 struct diff_filespec *two,
2874 const char *xfrm_msg,
2875 int complete_rewrite)
2876{
2877 const char *spawn_arg[10];
2878 int retval;
2879 const char **arg = &spawn_arg[0];
2880
2881 if (one && two) {
2882 struct diff_tempfile *temp_one, *temp_two;
2883 const char *othername = (other ? other : name);
2884 temp_one = prepare_temp_file(name, one);
2885 temp_two = prepare_temp_file(othername, two);
2886 *arg++ = pgm;
2887 *arg++ = name;
2888 *arg++ = temp_one->name;
2889 *arg++ = temp_one->hex;
2890 *arg++ = temp_one->mode;
2891 *arg++ = temp_two->name;
2892 *arg++ = temp_two->hex;
2893 *arg++ = temp_two->mode;
2894 if (other) {
2895 *arg++ = other;
2896 *arg++ = xfrm_msg;
2897 }
2898 } else {
2899 *arg++ = pgm;
2900 *arg++ = name;
2901 }
2902 *arg = NULL;
2903 fflush(NULL);
2904 retval = run_command_v_opt(spawn_arg, RUN_USING_SHELL);
2905 remove_tempfile();
2906 if (retval) {
2907 fprintf(stderr, "external diff died, stopping at %s.\n", name);
2908 exit(1);
2909 }
2910}
2911
2912static int similarity_index(struct diff_filepair *p)
2913{
2914 return p->score * 100 / MAX_SCORE;
2915}
2916
2917static void fill_metainfo(struct strbuf *msg,
2918 const char *name,
2919 const char *other,
2920 struct diff_filespec *one,
2921 struct diff_filespec *two,
2922 struct diff_options *o,
2923 struct diff_filepair *p,
2924 int *must_show_header,
2925 int use_color)
2926{
2927 const char *set = diff_get_color(use_color, DIFF_METAINFO);
2928 const char *reset = diff_get_color(use_color, DIFF_RESET);
2929 struct strbuf *msgbuf;
2930 char *line_prefix = "";
2931
2932 *must_show_header = 1;
2933 if (o->output_prefix) {
2934 msgbuf = o->output_prefix(o, o->output_prefix_data);
2935 line_prefix = msgbuf->buf;
2936 }
2937 strbuf_init(msg, PATH_MAX * 2 + 300);
2938 switch (p->status) {
2939 case DIFF_STATUS_COPIED:
2940 strbuf_addf(msg, "%s%ssimilarity index %d%%",
2941 line_prefix, set, similarity_index(p));
2942 strbuf_addf(msg, "%s\n%s%scopy from ",
2943 reset, line_prefix, set);
2944 quote_c_style(name, msg, NULL, 0);
2945 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
2946 quote_c_style(other, msg, NULL, 0);
2947 strbuf_addf(msg, "%s\n", reset);
2948 break;
2949 case DIFF_STATUS_RENAMED:
2950 strbuf_addf(msg, "%s%ssimilarity index %d%%",
2951 line_prefix, set, similarity_index(p));
2952 strbuf_addf(msg, "%s\n%s%srename from ",
2953 reset, line_prefix, set);
2954 quote_c_style(name, msg, NULL, 0);
2955 strbuf_addf(msg, "%s\n%s%srename to ",
2956 reset, line_prefix, set);
2957 quote_c_style(other, msg, NULL, 0);
2958 strbuf_addf(msg, "%s\n", reset);
2959 break;
2960 case DIFF_STATUS_MODIFIED:
2961 if (p->score) {
2962 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
2963 line_prefix,
2964 set, similarity_index(p), reset);
2965 break;
2966 }
2967 /* fallthru */
2968 default:
2969 *must_show_header = 0;
2970 }
2971 if (one && two && hashcmp(one->sha1, two->sha1)) {
2972 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2973
2974 if (DIFF_OPT_TST(o, BINARY)) {
2975 mmfile_t mf;
2976 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
2977 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
2978 abbrev = 40;
2979 }
2980 strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
2981 find_unique_abbrev(one->sha1, abbrev));
2982 strbuf_addstr(msg, find_unique_abbrev(two->sha1, abbrev));
2983 if (one->mode == two->mode)
2984 strbuf_addf(msg, " %06o", one->mode);
2985 strbuf_addf(msg, "%s\n", reset);
2986 }
2987}
2988
2989static void run_diff_cmd(const char *pgm,
2990 const char *name,
2991 const char *other,
2992 const char *attr_path,
2993 struct diff_filespec *one,
2994 struct diff_filespec *two,
2995 struct strbuf *msg,
2996 struct diff_options *o,
2997 struct diff_filepair *p)
2998{
2999 const char *xfrm_msg = NULL;
3000 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3001 int must_show_header = 0;
3002
3003 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3004 pgm = NULL;
3005 else {
3006 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3007 if (drv && drv->external)
3008 pgm = drv->external;
3009 }
3010
3011 if (msg) {
3012 /*
3013 * don't use colors when the header is intended for an
3014 * external diff driver
3015 */
3016 fill_metainfo(msg, name, other, one, two, o, p,
3017 &must_show_header,
3018 want_color(o->use_color) && !pgm);
3019 xfrm_msg = msg->len ? msg->buf : NULL;
3020 }
3021
3022 if (pgm) {
3023 run_external_diff(pgm, name, other, one, two, xfrm_msg,
3024 complete_rewrite);
3025 return;
3026 }
3027 if (one && two)
3028 builtin_diff(name, other ? other : name,
3029 one, two, xfrm_msg, must_show_header,
3030 o, complete_rewrite);
3031 else
3032 fprintf(o->file, "* Unmerged path %s\n", name);
3033}
3034
3035static void diff_fill_sha1_info(struct diff_filespec *one)
3036{
3037 if (DIFF_FILE_VALID(one)) {
3038 if (!one->sha1_valid) {
3039 struct stat st;
3040 if (!strcmp(one->path, "-")) {
3041 hashcpy(one->sha1, null_sha1);
3042 return;
3043 }
3044 if (lstat(one->path, &st) < 0)
3045 die_errno("stat '%s'", one->path);
3046 if (index_path(one->sha1, one->path, &st, 0))
3047 die("cannot hash %s", one->path);
3048 }
3049 }
3050 else
3051 hashclr(one->sha1);
3052}
3053
3054static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3055{
3056 /* Strip the prefix but do not molest /dev/null and absolute paths */
3057 if (*namep && **namep != '/') {
3058 *namep += prefix_length;
3059 if (**namep == '/')
3060 ++*namep;
3061 }
3062 if (*otherp && **otherp != '/') {
3063 *otherp += prefix_length;
3064 if (**otherp == '/')
3065 ++*otherp;
3066 }
3067}
3068
3069static void run_diff(struct diff_filepair *p, struct diff_options *o)
3070{
3071 const char *pgm = external_diff();
3072 struct strbuf msg;
3073 struct diff_filespec *one = p->one;
3074 struct diff_filespec *two = p->two;
3075 const char *name;
3076 const char *other;
3077 const char *attr_path;
3078
3079 name = p->one->path;
3080 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3081 attr_path = name;
3082 if (o->prefix_length)
3083 strip_prefix(o->prefix_length, &name, &other);
3084
3085 if (DIFF_PAIR_UNMERGED(p)) {
3086 run_diff_cmd(pgm, name, NULL, attr_path,
3087 NULL, NULL, NULL, o, p);
3088 return;
3089 }
3090
3091 diff_fill_sha1_info(one);
3092 diff_fill_sha1_info(two);
3093
3094 if (!pgm &&
3095 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3096 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3097 /*
3098 * a filepair that changes between file and symlink
3099 * needs to be split into deletion and creation.
3100 */
3101 struct diff_filespec *null = alloc_filespec(two->path);
3102 run_diff_cmd(NULL, name, other, attr_path,
3103 one, null, &msg, o, p);
3104 free(null);
3105 strbuf_release(&msg);
3106
3107 null = alloc_filespec(one->path);
3108 run_diff_cmd(NULL, name, other, attr_path,
3109 null, two, &msg, o, p);
3110 free(null);
3111 }
3112 else
3113 run_diff_cmd(pgm, name, other, attr_path,
3114 one, two, &msg, o, p);
3115
3116 strbuf_release(&msg);
3117}
3118
3119static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3120 struct diffstat_t *diffstat)
3121{
3122 const char *name;
3123 const char *other;
3124 int complete_rewrite = 0;
3125
3126 if (DIFF_PAIR_UNMERGED(p)) {
3127 /* unmerged */
3128 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
3129 return;
3130 }
3131
3132 name = p->one->path;
3133 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3134
3135 if (o->prefix_length)
3136 strip_prefix(o->prefix_length, &name, &other);
3137
3138 diff_fill_sha1_info(p->one);
3139 diff_fill_sha1_info(p->two);
3140
3141 if (p->status == DIFF_STATUS_MODIFIED && p->score)
3142 complete_rewrite = 1;
3143 builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
3144}
3145
3146static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3147{
3148 const char *name;
3149 const char *other;
3150 const char *attr_path;
3151
3152 if (DIFF_PAIR_UNMERGED(p)) {
3153 /* unmerged */
3154 return;
3155 }
3156
3157 name = p->one->path;
3158 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3159 attr_path = other ? other : name;
3160
3161 if (o->prefix_length)
3162 strip_prefix(o->prefix_length, &name, &other);
3163
3164 diff_fill_sha1_info(p->one);
3165 diff_fill_sha1_info(p->two);
3166
3167 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3168}
3169
3170void diff_setup(struct diff_options *options)
3171{
3172 memcpy(options, &default_diff_options, sizeof(*options));
3173
3174 options->file = stdout;
3175
3176 options->line_termination = '\n';
3177 options->break_opt = -1;
3178 options->rename_limit = -1;
3179 options->dirstat_permille = diff_dirstat_permille_default;
3180 options->context = 3;
3181 DIFF_OPT_SET(options, RENAME_EMPTY);
3182
3183 options->change = diff_change;
3184 options->add_remove = diff_addremove;
3185 options->use_color = diff_use_color_default;
3186 options->detect_rename = diff_detect_rename_default;
3187
3188 if (diff_no_prefix) {
3189 options->a_prefix = options->b_prefix = "";
3190 } else if (!diff_mnemonic_prefix) {
3191 options->a_prefix = "a/";
3192 options->b_prefix = "b/";
3193 }
3194}
3195
3196int diff_setup_done(struct diff_options *options)
3197{
3198 int count = 0;
3199
3200 if (options->output_format & DIFF_FORMAT_NAME)
3201 count++;
3202 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3203 count++;
3204 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3205 count++;
3206 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3207 count++;
3208 if (count > 1)
3209 die("--name-only, --name-status, --check and -s are mutually exclusive");
3210
3211 /*
3212 * Most of the time we can say "there are changes"
3213 * only by checking if there are changed paths, but
3214 * --ignore-whitespace* options force us to look
3215 * inside contents.
3216 */
3217
3218 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3219 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3220 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3221 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3222 else
3223 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3224
3225 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3226 options->detect_rename = DIFF_DETECT_COPY;
3227
3228 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3229 options->prefix = NULL;
3230 if (options->prefix)
3231 options->prefix_length = strlen(options->prefix);
3232 else
3233 options->prefix_length = 0;
3234
3235 if (options->output_format & (DIFF_FORMAT_NAME |
3236 DIFF_FORMAT_NAME_STATUS |
3237 DIFF_FORMAT_CHECKDIFF |
3238 DIFF_FORMAT_NO_OUTPUT))
3239 options->output_format &= ~(DIFF_FORMAT_RAW |
3240 DIFF_FORMAT_NUMSTAT |
3241 DIFF_FORMAT_DIFFSTAT |
3242 DIFF_FORMAT_SHORTSTAT |
3243 DIFF_FORMAT_DIRSTAT |
3244 DIFF_FORMAT_SUMMARY |
3245 DIFF_FORMAT_PATCH);
3246
3247 /*
3248 * These cases always need recursive; we do not drop caller-supplied
3249 * recursive bits for other formats here.
3250 */
3251 if (options->output_format & (DIFF_FORMAT_PATCH |
3252 DIFF_FORMAT_NUMSTAT |
3253 DIFF_FORMAT_DIFFSTAT |
3254 DIFF_FORMAT_SHORTSTAT |
3255 DIFF_FORMAT_DIRSTAT |
3256 DIFF_FORMAT_SUMMARY |
3257 DIFF_FORMAT_CHECKDIFF))
3258 DIFF_OPT_SET(options, RECURSIVE);
3259 /*
3260 * Also pickaxe would not work very well if you do not say recursive
3261 */
3262 if (options->pickaxe)
3263 DIFF_OPT_SET(options, RECURSIVE);
3264 /*
3265 * When patches are generated, submodules diffed against the work tree
3266 * must be checked for dirtiness too so it can be shown in the output
3267 */
3268 if (options->output_format & DIFF_FORMAT_PATCH)
3269 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3270
3271 if (options->detect_rename && options->rename_limit < 0)
3272 options->rename_limit = diff_rename_limit_default;
3273 if (options->setup & DIFF_SETUP_USE_CACHE) {
3274 if (!active_cache)
3275 /* read-cache does not die even when it fails
3276 * so it is safe for us to do this here. Also
3277 * it does not smudge active_cache or active_nr
3278 * when it fails, so we do not have to worry about
3279 * cleaning it up ourselves either.
3280 */
3281 read_cache();
3282 }
3283 if (options->abbrev <= 0 || 40 < options->abbrev)
3284 options->abbrev = 40; /* full */
3285
3286 /*
3287 * It does not make sense to show the first hit we happened
3288 * to have found. It does not make sense not to return with
3289 * exit code in such a case either.
3290 */
3291 if (DIFF_OPT_TST(options, QUICK)) {
3292 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3293 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3294 }
3295
3296 return 0;
3297}
3298
3299static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3300{
3301 char c, *eq;
3302 int len;
3303
3304 if (*arg != '-')
3305 return 0;
3306 c = *++arg;
3307 if (!c)
3308 return 0;
3309 if (c == arg_short) {
3310 c = *++arg;
3311 if (!c)
3312 return 1;
3313 if (val && isdigit(c)) {
3314 char *end;
3315 int n = strtoul(arg, &end, 10);
3316 if (*end)
3317 return 0;
3318 *val = n;
3319 return 1;
3320 }
3321 return 0;
3322 }
3323 if (c != '-')
3324 return 0;
3325 arg++;
3326 eq = strchr(arg, '=');
3327 if (eq)
3328 len = eq - arg;
3329 else
3330 len = strlen(arg);
3331 if (!len || strncmp(arg, arg_long, len))
3332 return 0;
3333 if (eq) {
3334 int n;
3335 char *end;
3336 if (!isdigit(*++eq))
3337 return 0;
3338 n = strtoul(eq, &end, 10);
3339 if (*end)
3340 return 0;
3341 *val = n;
3342 }
3343 return 1;
3344}
3345
3346static int diff_scoreopt_parse(const char *opt);
3347
3348static inline int short_opt(char opt, const char **argv,
3349 const char **optarg)
3350{
3351 const char *arg = argv[0];
3352 if (arg[0] != '-' || arg[1] != opt)
3353 return 0;
3354 if (arg[2] != '\0') {
3355 *optarg = arg + 2;
3356 return 1;
3357 }
3358 if (!argv[1])
3359 die("Option '%c' requires a value", opt);
3360 *optarg = argv[1];
3361 return 2;
3362}
3363
3364int parse_long_opt(const char *opt, const char **argv,
3365 const char **optarg)
3366{
3367 const char *arg = argv[0];
3368 if (arg[0] != '-' || arg[1] != '-')
3369 return 0;
3370 arg += strlen("--");
3371 if (prefixcmp(arg, opt))
3372 return 0;
3373 arg += strlen(opt);
3374 if (*arg == '=') { /* sticked form: --option=value */
3375 *optarg = arg + 1;
3376 return 1;
3377 }
3378 if (*arg != '\0')
3379 return 0;
3380 /* separate form: --option value */
3381 if (!argv[1])
3382 die("Option '--%s' requires a value", opt);
3383 *optarg = argv[1];
3384 return 2;
3385}
3386
3387static int stat_opt(struct diff_options *options, const char **av)
3388{
3389 const char *arg = av[0];
3390 char *end;
3391 int width = options->stat_width;
3392 int name_width = options->stat_name_width;
3393 int graph_width = options->stat_graph_width;
3394 int count = options->stat_count;
3395 int argcount = 1;
3396
3397 arg += strlen("--stat");
3398 end = (char *)arg;
3399
3400 switch (*arg) {
3401 case '-':
3402 if (!prefixcmp(arg, "-width")) {
3403 arg += strlen("-width");
3404 if (*arg == '=')
3405 width = strtoul(arg + 1, &end, 10);
3406 else if (!*arg && !av[1])
3407 die("Option '--stat-width' requires a value");
3408 else if (!*arg) {
3409 width = strtoul(av[1], &end, 10);
3410 argcount = 2;
3411 }
3412 } else if (!prefixcmp(arg, "-name-width")) {
3413 arg += strlen("-name-width");
3414 if (*arg == '=')
3415 name_width = strtoul(arg + 1, &end, 10);
3416 else if (!*arg && !av[1])
3417 die("Option '--stat-name-width' requires a value");
3418 else if (!*arg) {
3419 name_width = strtoul(av[1], &end, 10);
3420 argcount = 2;
3421 }
3422 } else if (!prefixcmp(arg, "-graph-width")) {
3423 arg += strlen("-graph-width");
3424 if (*arg == '=')
3425 graph_width = strtoul(arg + 1, &end, 10);
3426 else if (!*arg && !av[1])
3427 die("Option '--stat-graph-width' requires a value");
3428 else if (!*arg) {
3429 graph_width = strtoul(av[1], &end, 10);
3430 argcount = 2;
3431 }
3432 } else if (!prefixcmp(arg, "-count")) {
3433 arg += strlen("-count");
3434 if (*arg == '=')
3435 count = strtoul(arg + 1, &end, 10);
3436 else if (!*arg && !av[1])
3437 die("Option '--stat-count' requires a value");
3438 else if (!*arg) {
3439 count = strtoul(av[1], &end, 10);
3440 argcount = 2;
3441 }
3442 }
3443 break;
3444 case '=':
3445 width = strtoul(arg+1, &end, 10);
3446 if (*end == ',')
3447 name_width = strtoul(end+1, &end, 10);
3448 if (*end == ',')
3449 count = strtoul(end+1, &end, 10);
3450 }
3451
3452 /* Important! This checks all the error cases! */
3453 if (*end)
3454 return 0;
3455 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3456 options->stat_name_width = name_width;
3457 options->stat_graph_width = graph_width;
3458 options->stat_width = width;
3459 options->stat_count = count;
3460 return argcount;
3461}
3462
3463static int parse_dirstat_opt(struct diff_options *options, const char *params)
3464{
3465 struct strbuf errmsg = STRBUF_INIT;
3466 if (parse_dirstat_params(options, params, &errmsg))
3467 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3468 errmsg.buf);
3469 strbuf_release(&errmsg);
3470 /*
3471 * The caller knows a dirstat-related option is given from the command
3472 * line; allow it to say "return this_function();"
3473 */
3474 options->output_format |= DIFF_FORMAT_DIRSTAT;
3475 return 1;
3476}
3477
3478int diff_opt_parse(struct diff_options *options, const char **av, int ac)
3479{
3480 const char *arg = av[0];
3481 const char *optarg;
3482 int argcount;
3483
3484 /* Output format options */
3485 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch"))
3486 options->output_format |= DIFF_FORMAT_PATCH;
3487 else if (opt_arg(arg, 'U', "unified", &options->context))
3488 options->output_format |= DIFF_FORMAT_PATCH;
3489 else if (!strcmp(arg, "--raw"))
3490 options->output_format |= DIFF_FORMAT_RAW;
3491 else if (!strcmp(arg, "--patch-with-raw"))
3492 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
3493 else if (!strcmp(arg, "--numstat"))
3494 options->output_format |= DIFF_FORMAT_NUMSTAT;
3495 else if (!strcmp(arg, "--shortstat"))
3496 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3497 else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3498 return parse_dirstat_opt(options, "");
3499 else if (!prefixcmp(arg, "-X"))
3500 return parse_dirstat_opt(options, arg + 2);
3501 else if (!prefixcmp(arg, "--dirstat="))
3502 return parse_dirstat_opt(options, arg + 10);
3503 else if (!strcmp(arg, "--cumulative"))
3504 return parse_dirstat_opt(options, "cumulative");
3505 else if (!strcmp(arg, "--dirstat-by-file"))
3506 return parse_dirstat_opt(options, "files");
3507 else if (!prefixcmp(arg, "--dirstat-by-file=")) {
3508 parse_dirstat_opt(options, "files");
3509 return parse_dirstat_opt(options, arg + 18);
3510 }
3511 else if (!strcmp(arg, "--check"))
3512 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3513 else if (!strcmp(arg, "--summary"))
3514 options->output_format |= DIFF_FORMAT_SUMMARY;
3515 else if (!strcmp(arg, "--patch-with-stat"))
3516 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
3517 else if (!strcmp(arg, "--name-only"))
3518 options->output_format |= DIFF_FORMAT_NAME;
3519 else if (!strcmp(arg, "--name-status"))
3520 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3521 else if (!strcmp(arg, "-s"))
3522 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3523 else if (!prefixcmp(arg, "--stat"))
3524 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3525 return stat_opt(options, av);
3526
3527 /* renames options */
3528 else if (!prefixcmp(arg, "-B") || !prefixcmp(arg, "--break-rewrites=") ||
3529 !strcmp(arg, "--break-rewrites")) {
3530 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3531 return error("invalid argument to -B: %s", arg+2);
3532 }
3533 else if (!prefixcmp(arg, "-M") || !prefixcmp(arg, "--find-renames=") ||
3534 !strcmp(arg, "--find-renames")) {
3535 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3536 return error("invalid argument to -M: %s", arg+2);
3537 options->detect_rename = DIFF_DETECT_RENAME;
3538 }
3539 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3540 options->irreversible_delete = 1;
3541 }
3542 else if (!prefixcmp(arg, "-C") || !prefixcmp(arg, "--find-copies=") ||
3543 !strcmp(arg, "--find-copies")) {
3544 if (options->detect_rename == DIFF_DETECT_COPY)
3545 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3546 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3547 return error("invalid argument to -C: %s", arg+2);
3548 options->detect_rename = DIFF_DETECT_COPY;
3549 }
3550 else if (!strcmp(arg, "--no-renames"))
3551 options->detect_rename = 0;
3552 else if (!strcmp(arg, "--rename-empty"))
3553 DIFF_OPT_SET(options, RENAME_EMPTY);
3554 else if (!strcmp(arg, "--no-rename-empty"))
3555 DIFF_OPT_CLR(options, RENAME_EMPTY);
3556 else if (!strcmp(arg, "--relative"))
3557 DIFF_OPT_SET(options, RELATIVE_NAME);
3558 else if (!prefixcmp(arg, "--relative=")) {
3559 DIFF_OPT_SET(options, RELATIVE_NAME);
3560 options->prefix = arg + 11;
3561 }
3562
3563 /* xdiff options */
3564 else if (!strcmp(arg, "--minimal"))
3565 DIFF_XDL_SET(options, NEED_MINIMAL);
3566 else if (!strcmp(arg, "--no-minimal"))
3567 DIFF_XDL_CLR(options, NEED_MINIMAL);
3568 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3569 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3570 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3571 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3572 else if (!strcmp(arg, "--ignore-space-at-eol"))
3573 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3574 else if (!strcmp(arg, "--patience"))
3575 options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
3576 else if (!strcmp(arg, "--histogram"))
3577 options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
3578
3579 /* flags options */
3580 else if (!strcmp(arg, "--binary")) {
3581 options->output_format |= DIFF_FORMAT_PATCH;
3582 DIFF_OPT_SET(options, BINARY);
3583 }
3584 else if (!strcmp(arg, "--full-index"))
3585 DIFF_OPT_SET(options, FULL_INDEX);
3586 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3587 DIFF_OPT_SET(options, TEXT);
3588 else if (!strcmp(arg, "-R"))
3589 DIFF_OPT_SET(options, REVERSE_DIFF);
3590 else if (!strcmp(arg, "--find-copies-harder"))
3591 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3592 else if (!strcmp(arg, "--follow"))
3593 DIFF_OPT_SET(options, FOLLOW_RENAMES);
3594 else if (!strcmp(arg, "--color"))
3595 options->use_color = 1;
3596 else if (!prefixcmp(arg, "--color=")) {
3597 int value = git_config_colorbool(NULL, arg+8);
3598 if (value < 0)
3599 return error("option `color' expects \"always\", \"auto\", or \"never\"");
3600 options->use_color = value;
3601 }
3602 else if (!strcmp(arg, "--no-color"))
3603 options->use_color = 0;
3604 else if (!strcmp(arg, "--color-words")) {
3605 options->use_color = 1;
3606 options->word_diff = DIFF_WORDS_COLOR;
3607 }
3608 else if (!prefixcmp(arg, "--color-words=")) {
3609 options->use_color = 1;
3610 options->word_diff = DIFF_WORDS_COLOR;
3611 options->word_regex = arg + 14;
3612 }
3613 else if (!strcmp(arg, "--word-diff")) {
3614 if (options->word_diff == DIFF_WORDS_NONE)
3615 options->word_diff = DIFF_WORDS_PLAIN;
3616 }
3617 else if (!prefixcmp(arg, "--word-diff=")) {
3618 const char *type = arg + 12;
3619 if (!strcmp(type, "plain"))
3620 options->word_diff = DIFF_WORDS_PLAIN;
3621 else if (!strcmp(type, "color")) {
3622 options->use_color = 1;
3623 options->word_diff = DIFF_WORDS_COLOR;
3624 }
3625 else if (!strcmp(type, "porcelain"))
3626 options->word_diff = DIFF_WORDS_PORCELAIN;
3627 else if (!strcmp(type, "none"))
3628 options->word_diff = DIFF_WORDS_NONE;
3629 else
3630 die("bad --word-diff argument: %s", type);
3631 }
3632 else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3633 if (options->word_diff == DIFF_WORDS_NONE)
3634 options->word_diff = DIFF_WORDS_PLAIN;
3635 options->word_regex = optarg;
3636 return argcount;
3637 }
3638 else if (!strcmp(arg, "--exit-code"))
3639 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3640 else if (!strcmp(arg, "--quiet"))
3641 DIFF_OPT_SET(options, QUICK);
3642 else if (!strcmp(arg, "--ext-diff"))
3643 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3644 else if (!strcmp(arg, "--no-ext-diff"))
3645 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3646 else if (!strcmp(arg, "--textconv"))
3647 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3648 else if (!strcmp(arg, "--no-textconv"))
3649 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3650 else if (!strcmp(arg, "--ignore-submodules")) {
3651 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3652 handle_ignore_submodules_arg(options, "all");
3653 } else if (!prefixcmp(arg, "--ignore-submodules=")) {
3654 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3655 handle_ignore_submodules_arg(options, arg + 20);
3656 } else if (!strcmp(arg, "--submodule"))
3657 DIFF_OPT_SET(options, SUBMODULE_LOG);
3658 else if (!prefixcmp(arg, "--submodule=")) {
3659 if (!strcmp(arg + 12, "log"))
3660 DIFF_OPT_SET(options, SUBMODULE_LOG);
3661 }
3662
3663 /* misc options */
3664 else if (!strcmp(arg, "-z"))
3665 options->line_termination = 0;
3666 else if ((argcount = short_opt('l', av, &optarg))) {
3667 options->rename_limit = strtoul(optarg, NULL, 10);
3668 return argcount;
3669 }
3670 else if ((argcount = short_opt('S', av, &optarg))) {
3671 options->pickaxe = optarg;
3672 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3673 return argcount;
3674 } else if ((argcount = short_opt('G', av, &optarg))) {
3675 options->pickaxe = optarg;
3676 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3677 return argcount;
3678 }
3679 else if (!strcmp(arg, "--pickaxe-all"))
3680 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3681 else if (!strcmp(arg, "--pickaxe-regex"))
3682 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3683 else if ((argcount = short_opt('O', av, &optarg))) {
3684 options->orderfile = optarg;
3685 return argcount;
3686 }
3687 else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3688 options->filter = optarg;
3689 return argcount;
3690 }
3691 else if (!strcmp(arg, "--abbrev"))
3692 options->abbrev = DEFAULT_ABBREV;
3693 else if (!prefixcmp(arg, "--abbrev=")) {
3694 options->abbrev = strtoul(arg + 9, NULL, 10);
3695 if (options->abbrev < MINIMUM_ABBREV)
3696 options->abbrev = MINIMUM_ABBREV;
3697 else if (40 < options->abbrev)
3698 options->abbrev = 40;
3699 }
3700 else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3701 options->a_prefix = optarg;
3702 return argcount;
3703 }
3704 else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
3705 options->b_prefix = optarg;
3706 return argcount;
3707 }
3708 else if (!strcmp(arg, "--no-prefix"))
3709 options->a_prefix = options->b_prefix = "";
3710 else if (opt_arg(arg, '\0', "inter-hunk-context",
3711 &options->interhunkcontext))
3712 ;
3713 else if (!strcmp(arg, "-W"))
3714 DIFF_OPT_SET(options, FUNCCONTEXT);
3715 else if (!strcmp(arg, "--function-context"))
3716 DIFF_OPT_SET(options, FUNCCONTEXT);
3717 else if (!strcmp(arg, "--no-function-context"))
3718 DIFF_OPT_CLR(options, FUNCCONTEXT);
3719 else if ((argcount = parse_long_opt("output", av, &optarg))) {
3720 options->file = fopen(optarg, "w");
3721 if (!options->file)
3722 die_errno("Could not open '%s'", optarg);
3723 options->close_file = 1;
3724 return argcount;
3725 } else
3726 return 0;
3727 return 1;
3728}
3729
3730int parse_rename_score(const char **cp_p)
3731{
3732 unsigned long num, scale;
3733 int ch, dot;
3734 const char *cp = *cp_p;
3735
3736 num = 0;
3737 scale = 1;
3738 dot = 0;
3739 for (;;) {
3740 ch = *cp;
3741 if ( !dot && ch == '.' ) {
3742 scale = 1;
3743 dot = 1;
3744 } else if ( ch == '%' ) {
3745 scale = dot ? scale*100 : 100;
3746 cp++; /* % is always at the end */
3747 break;
3748 } else if ( ch >= '0' && ch <= '9' ) {
3749 if ( scale < 100000 ) {
3750 scale *= 10;
3751 num = (num*10) + (ch-'0');
3752 }
3753 } else {
3754 break;
3755 }
3756 cp++;
3757 }
3758 *cp_p = cp;
3759
3760 /* user says num divided by scale and we say internally that
3761 * is MAX_SCORE * num / scale.
3762 */
3763 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
3764}
3765
3766static int diff_scoreopt_parse(const char *opt)
3767{
3768 int opt1, opt2, cmd;
3769
3770 if (*opt++ != '-')
3771 return -1;
3772 cmd = *opt++;
3773 if (cmd == '-') {
3774 /* convert the long-form arguments into short-form versions */
3775 if (!prefixcmp(opt, "break-rewrites")) {
3776 opt += strlen("break-rewrites");
3777 if (*opt == 0 || *opt++ == '=')
3778 cmd = 'B';
3779 } else if (!prefixcmp(opt, "find-copies")) {
3780 opt += strlen("find-copies");
3781 if (*opt == 0 || *opt++ == '=')
3782 cmd = 'C';
3783 } else if (!prefixcmp(opt, "find-renames")) {
3784 opt += strlen("find-renames");
3785 if (*opt == 0 || *opt++ == '=')
3786 cmd = 'M';
3787 }
3788 }
3789 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
3790 return -1; /* that is not a -M, -C nor -B option */
3791
3792 opt1 = parse_rename_score(&opt);
3793 if (cmd != 'B')
3794 opt2 = 0;
3795 else {
3796 if (*opt == 0)
3797 opt2 = 0;
3798 else if (*opt != '/')
3799 return -1; /* we expect -B80/99 or -B80 */
3800 else {
3801 opt++;
3802 opt2 = parse_rename_score(&opt);
3803 }
3804 }
3805 if (*opt != 0)
3806 return -1;
3807 return opt1 | (opt2 << 16);
3808}
3809
3810struct diff_queue_struct diff_queued_diff;
3811
3812void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
3813{
3814 if (queue->alloc <= queue->nr) {
3815 queue->alloc = alloc_nr(queue->alloc);
3816 queue->queue = xrealloc(queue->queue,
3817 sizeof(dp) * queue->alloc);
3818 }
3819 queue->queue[queue->nr++] = dp;
3820}
3821
3822struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
3823 struct diff_filespec *one,
3824 struct diff_filespec *two)
3825{
3826 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
3827 dp->one = one;
3828 dp->two = two;
3829 if (queue)
3830 diff_q(queue, dp);
3831 return dp;
3832}
3833
3834void diff_free_filepair(struct diff_filepair *p)
3835{
3836 free_filespec(p->one);
3837 free_filespec(p->two);
3838 free(p);
3839}
3840
3841/* This is different from find_unique_abbrev() in that
3842 * it stuffs the result with dots for alignment.
3843 */
3844const char *diff_unique_abbrev(const unsigned char *sha1, int len)
3845{
3846 int abblen;
3847 const char *abbrev;
3848 if (len == 40)
3849 return sha1_to_hex(sha1);
3850
3851 abbrev = find_unique_abbrev(sha1, len);
3852 abblen = strlen(abbrev);
3853 if (abblen < 37) {
3854 static char hex[41];
3855 if (len < abblen && abblen <= len + 2)
3856 sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
3857 else
3858 sprintf(hex, "%s...", abbrev);
3859 return hex;
3860 }
3861 return sha1_to_hex(sha1);
3862}
3863
3864static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
3865{
3866 int line_termination = opt->line_termination;
3867 int inter_name_termination = line_termination ? '\t' : '\0';
3868 if (opt->output_prefix) {
3869 struct strbuf *msg = NULL;
3870 msg = opt->output_prefix(opt, opt->output_prefix_data);
3871 fprintf(opt->file, "%s", msg->buf);
3872 }
3873
3874 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
3875 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
3876 diff_unique_abbrev(p->one->sha1, opt->abbrev));
3877 fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
3878 }
3879 if (p->score) {
3880 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
3881 inter_name_termination);
3882 } else {
3883 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
3884 }
3885
3886 if (p->status == DIFF_STATUS_COPIED ||
3887 p->status == DIFF_STATUS_RENAMED) {
3888 const char *name_a, *name_b;
3889 name_a = p->one->path;
3890 name_b = p->two->path;
3891 strip_prefix(opt->prefix_length, &name_a, &name_b);
3892 write_name_quoted(name_a, opt->file, inter_name_termination);
3893 write_name_quoted(name_b, opt->file, line_termination);
3894 } else {
3895 const char *name_a, *name_b;
3896 name_a = p->one->mode ? p->one->path : p->two->path;
3897 name_b = NULL;
3898 strip_prefix(opt->prefix_length, &name_a, &name_b);
3899 write_name_quoted(name_a, opt->file, line_termination);
3900 }
3901}
3902
3903int diff_unmodified_pair(struct diff_filepair *p)
3904{
3905 /* This function is written stricter than necessary to support
3906 * the currently implemented transformers, but the idea is to
3907 * let transformers to produce diff_filepairs any way they want,
3908 * and filter and clean them up here before producing the output.
3909 */
3910 struct diff_filespec *one = p->one, *two = p->two;
3911
3912 if (DIFF_PAIR_UNMERGED(p))
3913 return 0; /* unmerged is interesting */
3914
3915 /* deletion, addition, mode or type change
3916 * and rename are all interesting.
3917 */
3918 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
3919 DIFF_PAIR_MODE_CHANGED(p) ||
3920 strcmp(one->path, two->path))
3921 return 0;
3922
3923 /* both are valid and point at the same path. that is, we are
3924 * dealing with a change.
3925 */
3926 if (one->sha1_valid && two->sha1_valid &&
3927 !hashcmp(one->sha1, two->sha1) &&
3928 !one->dirty_submodule && !two->dirty_submodule)
3929 return 1; /* no change */
3930 if (!one->sha1_valid && !two->sha1_valid)
3931 return 1; /* both look at the same file on the filesystem. */
3932 return 0;
3933}
3934
3935static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
3936{
3937 if (diff_unmodified_pair(p))
3938 return;
3939
3940 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3941 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3942 return; /* no tree diffs in patch format */
3943
3944 run_diff(p, o);
3945}
3946
3947static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
3948 struct diffstat_t *diffstat)
3949{
3950 if (diff_unmodified_pair(p))
3951 return;
3952
3953 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3954 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3955 return; /* no useful stat for tree diffs */
3956
3957 run_diffstat(p, o, diffstat);
3958}
3959
3960static void diff_flush_checkdiff(struct diff_filepair *p,
3961 struct diff_options *o)
3962{
3963 if (diff_unmodified_pair(p))
3964 return;
3965
3966 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3967 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3968 return; /* nothing to check in tree diffs */
3969
3970 run_checkdiff(p, o);
3971}
3972
3973int diff_queue_is_empty(void)
3974{
3975 struct diff_queue_struct *q = &diff_queued_diff;
3976 int i;
3977 for (i = 0; i < q->nr; i++)
3978 if (!diff_unmodified_pair(q->queue[i]))
3979 return 0;
3980 return 1;
3981}
3982
3983#if DIFF_DEBUG
3984void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
3985{
3986 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
3987 x, one ? one : "",
3988 s->path,
3989 DIFF_FILE_VALID(s) ? "valid" : "invalid",
3990 s->mode,
3991 s->sha1_valid ? sha1_to_hex(s->sha1) : "");
3992 fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
3993 x, one ? one : "",
3994 s->size, s->xfrm_flags);
3995}
3996
3997void diff_debug_filepair(const struct diff_filepair *p, int i)
3998{
3999 diff_debug_filespec(p->one, i, "one");
4000 diff_debug_filespec(p->two, i, "two");
4001 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4002 p->score, p->status ? p->status : '?',
4003 p->one->rename_used, p->broken_pair);
4004}
4005
4006void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4007{
4008 int i;
4009 if (msg)
4010 fprintf(stderr, "%s\n", msg);
4011 fprintf(stderr, "q->nr = %d\n", q->nr);
4012 for (i = 0; i < q->nr; i++) {
4013 struct diff_filepair *p = q->queue[i];
4014 diff_debug_filepair(p, i);
4015 }
4016}
4017#endif
4018
4019static void diff_resolve_rename_copy(void)
4020{
4021 int i;
4022 struct diff_filepair *p;
4023 struct diff_queue_struct *q = &diff_queued_diff;
4024
4025 diff_debug_queue("resolve-rename-copy", q);
4026
4027 for (i = 0; i < q->nr; i++) {
4028 p = q->queue[i];
4029 p->status = 0; /* undecided */
4030 if (DIFF_PAIR_UNMERGED(p))
4031 p->status = DIFF_STATUS_UNMERGED;
4032 else if (!DIFF_FILE_VALID(p->one))
4033 p->status = DIFF_STATUS_ADDED;
4034 else if (!DIFF_FILE_VALID(p->two))
4035 p->status = DIFF_STATUS_DELETED;
4036 else if (DIFF_PAIR_TYPE_CHANGED(p))
4037 p->status = DIFF_STATUS_TYPE_CHANGED;
4038
4039 /* from this point on, we are dealing with a pair
4040 * whose both sides are valid and of the same type, i.e.
4041 * either in-place edit or rename/copy edit.
4042 */
4043 else if (DIFF_PAIR_RENAME(p)) {
4044 /*
4045 * A rename might have re-connected a broken
4046 * pair up, causing the pathnames to be the
4047 * same again. If so, that's not a rename at
4048 * all, just a modification..
4049 *
4050 * Otherwise, see if this source was used for
4051 * multiple renames, in which case we decrement
4052 * the count, and call it a copy.
4053 */
4054 if (!strcmp(p->one->path, p->two->path))
4055 p->status = DIFF_STATUS_MODIFIED;
4056 else if (--p->one->rename_used > 0)
4057 p->status = DIFF_STATUS_COPIED;
4058 else
4059 p->status = DIFF_STATUS_RENAMED;
4060 }
4061 else if (hashcmp(p->one->sha1, p->two->sha1) ||
4062 p->one->mode != p->two->mode ||
4063 p->one->dirty_submodule ||
4064 p->two->dirty_submodule ||
4065 is_null_sha1(p->one->sha1))
4066 p->status = DIFF_STATUS_MODIFIED;
4067 else {
4068 /* This is a "no-change" entry and should not
4069 * happen anymore, but prepare for broken callers.
4070 */
4071 error("feeding unmodified %s to diffcore",
4072 p->one->path);
4073 p->status = DIFF_STATUS_UNKNOWN;
4074 }
4075 }
4076 diff_debug_queue("resolve-rename-copy done", q);
4077}
4078
4079static int check_pair_status(struct diff_filepair *p)
4080{
4081 switch (p->status) {
4082 case DIFF_STATUS_UNKNOWN:
4083 return 0;
4084 case 0:
4085 die("internal error in diff-resolve-rename-copy");
4086 default:
4087 return 1;
4088 }
4089}
4090
4091static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4092{
4093 int fmt = opt->output_format;
4094
4095 if (fmt & DIFF_FORMAT_CHECKDIFF)
4096 diff_flush_checkdiff(p, opt);
4097 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4098 diff_flush_raw(p, opt);
4099 else if (fmt & DIFF_FORMAT_NAME) {
4100 const char *name_a, *name_b;
4101 name_a = p->two->path;
4102 name_b = NULL;
4103 strip_prefix(opt->prefix_length, &name_a, &name_b);
4104 write_name_quoted(name_a, opt->file, opt->line_termination);
4105 }
4106}
4107
4108static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4109{
4110 if (fs->mode)
4111 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4112 else
4113 fprintf(file, " %s ", newdelete);
4114 write_name_quoted(fs->path, file, '\n');
4115}
4116
4117
4118static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4119 const char *line_prefix)
4120{
4121 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4122 fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4123 p->two->mode, show_name ? ' ' : '\n');
4124 if (show_name) {
4125 write_name_quoted(p->two->path, file, '\n');
4126 }
4127 }
4128}
4129
4130static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4131 const char *line_prefix)
4132{
4133 char *names = pprint_rename(p->one->path, p->two->path);
4134
4135 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4136 free(names);
4137 show_mode_change(file, p, 0, line_prefix);
4138}
4139
4140static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4141{
4142 FILE *file = opt->file;
4143 char *line_prefix = "";
4144
4145 if (opt->output_prefix) {
4146 struct strbuf *buf = opt->output_prefix(opt, opt->output_prefix_data);
4147 line_prefix = buf->buf;
4148 }
4149
4150 switch(p->status) {
4151 case DIFF_STATUS_DELETED:
4152 fputs(line_prefix, file);
4153 show_file_mode_name(file, "delete", p->one);
4154 break;
4155 case DIFF_STATUS_ADDED:
4156 fputs(line_prefix, file);
4157 show_file_mode_name(file, "create", p->two);
4158 break;
4159 case DIFF_STATUS_COPIED:
4160 fputs(line_prefix, file);
4161 show_rename_copy(file, "copy", p, line_prefix);
4162 break;
4163 case DIFF_STATUS_RENAMED:
4164 fputs(line_prefix, file);
4165 show_rename_copy(file, "rename", p, line_prefix);
4166 break;
4167 default:
4168 if (p->score) {
4169 fprintf(file, "%s rewrite ", line_prefix);
4170 write_name_quoted(p->two->path, file, ' ');
4171 fprintf(file, "(%d%%)\n", similarity_index(p));
4172 }
4173 show_mode_change(file, p, !p->score, line_prefix);
4174 break;
4175 }
4176}
4177
4178struct patch_id_t {
4179 git_SHA_CTX *ctx;
4180 int patchlen;
4181};
4182
4183static int remove_space(char *line, int len)
4184{
4185 int i;
4186 char *dst = line;
4187 unsigned char c;
4188
4189 for (i = 0; i < len; i++)
4190 if (!isspace((c = line[i])))
4191 *dst++ = c;
4192
4193 return dst - line;
4194}
4195
4196static void patch_id_consume(void *priv, char *line, unsigned long len)
4197{
4198 struct patch_id_t *data = priv;
4199 int new_len;
4200
4201 /* Ignore line numbers when computing the SHA1 of the patch */
4202 if (!prefixcmp(line, "@@ -"))
4203 return;
4204
4205 new_len = remove_space(line, len);
4206
4207 git_SHA1_Update(data->ctx, line, new_len);
4208 data->patchlen += new_len;
4209}
4210
4211/* returns 0 upon success, and writes result into sha1 */
4212static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
4213{
4214 struct diff_queue_struct *q = &diff_queued_diff;
4215 int i;
4216 git_SHA_CTX ctx;
4217 struct patch_id_t data;
4218 char buffer[PATH_MAX * 4 + 20];
4219
4220 git_SHA1_Init(&ctx);
4221 memset(&data, 0, sizeof(struct patch_id_t));
4222 data.ctx = &ctx;
4223
4224 for (i = 0; i < q->nr; i++) {
4225 xpparam_t xpp;
4226 xdemitconf_t xecfg;
4227 mmfile_t mf1, mf2;
4228 struct diff_filepair *p = q->queue[i];
4229 int len1, len2;
4230
4231 memset(&xpp, 0, sizeof(xpp));
4232 memset(&xecfg, 0, sizeof(xecfg));
4233 if (p->status == 0)
4234 return error("internal diff status error");
4235 if (p->status == DIFF_STATUS_UNKNOWN)
4236 continue;
4237 if (diff_unmodified_pair(p))
4238 continue;
4239 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4240 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4241 continue;
4242 if (DIFF_PAIR_UNMERGED(p))
4243 continue;
4244
4245 diff_fill_sha1_info(p->one);
4246 diff_fill_sha1_info(p->two);
4247 if (fill_mmfile(&mf1, p->one) < 0 ||
4248 fill_mmfile(&mf2, p->two) < 0)
4249 return error("unable to read files to diff");
4250
4251 len1 = remove_space(p->one->path, strlen(p->one->path));
4252 len2 = remove_space(p->two->path, strlen(p->two->path));
4253 if (p->one->mode == 0)
4254 len1 = snprintf(buffer, sizeof(buffer),
4255 "diff--gita/%.*sb/%.*s"
4256 "newfilemode%06o"
4257 "---/dev/null"
4258 "+++b/%.*s",
4259 len1, p->one->path,
4260 len2, p->two->path,
4261 p->two->mode,
4262 len2, p->two->path);
4263 else if (p->two->mode == 0)
4264 len1 = snprintf(buffer, sizeof(buffer),
4265 "diff--gita/%.*sb/%.*s"
4266 "deletedfilemode%06o"
4267 "---a/%.*s"
4268 "+++/dev/null",
4269 len1, p->one->path,
4270 len2, p->two->path,
4271 p->one->mode,
4272 len1, p->one->path);
4273 else
4274 len1 = snprintf(buffer, sizeof(buffer),
4275 "diff--gita/%.*sb/%.*s"
4276 "---a/%.*s"
4277 "+++b/%.*s",
4278 len1, p->one->path,
4279 len2, p->two->path,
4280 len1, p->one->path,
4281 len2, p->two->path);
4282 git_SHA1_Update(&ctx, buffer, len1);
4283
4284 if (diff_filespec_is_binary(p->one) ||
4285 diff_filespec_is_binary(p->two)) {
4286 git_SHA1_Update(&ctx, sha1_to_hex(p->one->sha1), 40);
4287 git_SHA1_Update(&ctx, sha1_to_hex(p->two->sha1), 40);
4288 continue;
4289 }
4290
4291 xpp.flags = 0;
4292 xecfg.ctxlen = 3;
4293 xecfg.flags = 0;
4294 xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4295 &xpp, &xecfg);
4296 }
4297
4298 git_SHA1_Final(sha1, &ctx);
4299 return 0;
4300}
4301
4302int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
4303{
4304 struct diff_queue_struct *q = &diff_queued_diff;
4305 int i;
4306 int result = diff_get_patch_id(options, sha1);
4307
4308 for (i = 0; i < q->nr; i++)
4309 diff_free_filepair(q->queue[i]);
4310
4311 free(q->queue);
4312 DIFF_QUEUE_CLEAR(q);
4313
4314 return result;
4315}
4316
4317static int is_summary_empty(const struct diff_queue_struct *q)
4318{
4319 int i;
4320
4321 for (i = 0; i < q->nr; i++) {
4322 const struct diff_filepair *p = q->queue[i];
4323
4324 switch (p->status) {
4325 case DIFF_STATUS_DELETED:
4326 case DIFF_STATUS_ADDED:
4327 case DIFF_STATUS_COPIED:
4328 case DIFF_STATUS_RENAMED:
4329 return 0;
4330 default:
4331 if (p->score)
4332 return 0;
4333 if (p->one->mode && p->two->mode &&
4334 p->one->mode != p->two->mode)
4335 return 0;
4336 break;
4337 }
4338 }
4339 return 1;
4340}
4341
4342static const char rename_limit_warning[] =
4343"inexact rename detection was skipped due to too many files.";
4344
4345static const char degrade_cc_to_c_warning[] =
4346"only found copies from modified paths due to too many files.";
4347
4348static const char rename_limit_advice[] =
4349"you may want to set your %s variable to at least "
4350"%d and retry the command.";
4351
4352void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4353{
4354 if (degraded_cc)
4355 warning(degrade_cc_to_c_warning);
4356 else if (needed)
4357 warning(rename_limit_warning);
4358 else
4359 return;
4360 if (0 < needed && needed < 32767)
4361 warning(rename_limit_advice, varname, needed);
4362}
4363
4364void diff_flush(struct diff_options *options)
4365{
4366 struct diff_queue_struct *q = &diff_queued_diff;
4367 int i, output_format = options->output_format;
4368 int separator = 0;
4369 int dirstat_by_line = 0;
4370
4371 /*
4372 * Order: raw, stat, summary, patch
4373 * or: name/name-status/checkdiff (other bits clear)
4374 */
4375 if (!q->nr)
4376 goto free_queue;
4377
4378 if (output_format & (DIFF_FORMAT_RAW |
4379 DIFF_FORMAT_NAME |
4380 DIFF_FORMAT_NAME_STATUS |
4381 DIFF_FORMAT_CHECKDIFF)) {
4382 for (i = 0; i < q->nr; i++) {
4383 struct diff_filepair *p = q->queue[i];
4384 if (check_pair_status(p))
4385 flush_one_pair(p, options);
4386 }
4387 separator++;
4388 }
4389
4390 if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4391 dirstat_by_line = 1;
4392
4393 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4394 dirstat_by_line) {
4395 struct diffstat_t diffstat;
4396
4397 memset(&diffstat, 0, sizeof(struct diffstat_t));
4398 for (i = 0; i < q->nr; i++) {
4399 struct diff_filepair *p = q->queue[i];
4400 if (check_pair_status(p))
4401 diff_flush_stat(p, options, &diffstat);
4402 }
4403 if (output_format & DIFF_FORMAT_NUMSTAT)
4404 show_numstat(&diffstat, options);
4405 if (output_format & DIFF_FORMAT_DIFFSTAT)
4406 show_stats(&diffstat, options);
4407 if (output_format & DIFF_FORMAT_SHORTSTAT)
4408 show_shortstats(&diffstat, options);
4409 if (output_format & DIFF_FORMAT_DIRSTAT)
4410 show_dirstat_by_line(&diffstat, options);
4411 free_diffstat_info(&diffstat);
4412 separator++;
4413 }
4414 if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4415 show_dirstat(options);
4416
4417 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4418 for (i = 0; i < q->nr; i++) {
4419 diff_summary(options, q->queue[i]);
4420 }
4421 separator++;
4422 }
4423
4424 if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4425 DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4426 DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4427 /*
4428 * run diff_flush_patch for the exit status. setting
4429 * options->file to /dev/null should be safe, becaue we
4430 * aren't supposed to produce any output anyway.
4431 */
4432 if (options->close_file)
4433 fclose(options->file);
4434 options->file = fopen("/dev/null", "w");
4435 if (!options->file)
4436 die_errno("Could not open /dev/null");
4437 options->close_file = 1;
4438 for (i = 0; i < q->nr; i++) {
4439 struct diff_filepair *p = q->queue[i];
4440 if (check_pair_status(p))
4441 diff_flush_patch(p, options);
4442 if (options->found_changes)
4443 break;
4444 }
4445 }
4446
4447 if (output_format & DIFF_FORMAT_PATCH) {
4448 if (separator) {
4449 if (options->output_prefix) {
4450 struct strbuf *msg = NULL;
4451 msg = options->output_prefix(options,
4452 options->output_prefix_data);
4453 fwrite(msg->buf, msg->len, 1, stdout);
4454 }
4455 putc(options->line_termination, options->file);
4456 if (options->stat_sep) {
4457 /* attach patch instead of inline */
4458 fputs(options->stat_sep, options->file);
4459 }
4460 }
4461
4462 for (i = 0; i < q->nr; i++) {
4463 struct diff_filepair *p = q->queue[i];
4464 if (check_pair_status(p))
4465 diff_flush_patch(p, options);
4466 }
4467 }
4468
4469 if (output_format & DIFF_FORMAT_CALLBACK)
4470 options->format_callback(q, options, options->format_callback_data);
4471
4472 for (i = 0; i < q->nr; i++)
4473 diff_free_filepair(q->queue[i]);
4474free_queue:
4475 free(q->queue);
4476 DIFF_QUEUE_CLEAR(q);
4477 if (options->close_file)
4478 fclose(options->file);
4479
4480 /*
4481 * Report the content-level differences with HAS_CHANGES;
4482 * diff_addremove/diff_change does not set the bit when
4483 * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4484 */
4485 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4486 if (options->found_changes)
4487 DIFF_OPT_SET(options, HAS_CHANGES);
4488 else
4489 DIFF_OPT_CLR(options, HAS_CHANGES);
4490 }
4491}
4492
4493static void diffcore_apply_filter(const char *filter)
4494{
4495 int i;
4496 struct diff_queue_struct *q = &diff_queued_diff;
4497 struct diff_queue_struct outq;
4498 DIFF_QUEUE_CLEAR(&outq);
4499
4500 if (!filter)
4501 return;
4502
4503 if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
4504 int found;
4505 for (i = found = 0; !found && i < q->nr; i++) {
4506 struct diff_filepair *p = q->queue[i];
4507 if (((p->status == DIFF_STATUS_MODIFIED) &&
4508 ((p->score &&
4509 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4510 (!p->score &&
4511 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4512 ((p->status != DIFF_STATUS_MODIFIED) &&
4513 strchr(filter, p->status)))
4514 found++;
4515 }
4516 if (found)
4517 return;
4518
4519 /* otherwise we will clear the whole queue
4520 * by copying the empty outq at the end of this
4521 * function, but first clear the current entries
4522 * in the queue.
4523 */
4524 for (i = 0; i < q->nr; i++)
4525 diff_free_filepair(q->queue[i]);
4526 }
4527 else {
4528 /* Only the matching ones */
4529 for (i = 0; i < q->nr; i++) {
4530 struct diff_filepair *p = q->queue[i];
4531
4532 if (((p->status == DIFF_STATUS_MODIFIED) &&
4533 ((p->score &&
4534 strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4535 (!p->score &&
4536 strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4537 ((p->status != DIFF_STATUS_MODIFIED) &&
4538 strchr(filter, p->status)))
4539 diff_q(&outq, p);
4540 else
4541 diff_free_filepair(p);
4542 }
4543 }
4544 free(q->queue);
4545 *q = outq;
4546}
4547
4548/* Check whether two filespecs with the same mode and size are identical */
4549static int diff_filespec_is_identical(struct diff_filespec *one,
4550 struct diff_filespec *two)
4551{
4552 if (S_ISGITLINK(one->mode))
4553 return 0;
4554 if (diff_populate_filespec(one, 0))
4555 return 0;
4556 if (diff_populate_filespec(two, 0))
4557 return 0;
4558 return !memcmp(one->data, two->data, one->size);
4559}
4560
4561static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4562{
4563 int i;
4564 struct diff_queue_struct *q = &diff_queued_diff;
4565 struct diff_queue_struct outq;
4566 DIFF_QUEUE_CLEAR(&outq);
4567
4568 for (i = 0; i < q->nr; i++) {
4569 struct diff_filepair *p = q->queue[i];
4570
4571 /*
4572 * 1. Entries that come from stat info dirtiness
4573 * always have both sides (iow, not create/delete),
4574 * one side of the object name is unknown, with
4575 * the same mode and size. Keep the ones that
4576 * do not match these criteria. They have real
4577 * differences.
4578 *
4579 * 2. At this point, the file is known to be modified,
4580 * with the same mode and size, and the object
4581 * name of one side is unknown. Need to inspect
4582 * the identical contents.
4583 */
4584 if (!DIFF_FILE_VALID(p->one) || /* (1) */
4585 !DIFF_FILE_VALID(p->two) ||
4586 (p->one->sha1_valid && p->two->sha1_valid) ||
4587 (p->one->mode != p->two->mode) ||
4588 diff_populate_filespec(p->one, 1) ||
4589 diff_populate_filespec(p->two, 1) ||
4590 (p->one->size != p->two->size) ||
4591 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4592 diff_q(&outq, p);
4593 else {
4594 /*
4595 * The caller can subtract 1 from skip_stat_unmatch
4596 * to determine how many paths were dirty only
4597 * due to stat info mismatch.
4598 */
4599 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4600 diffopt->skip_stat_unmatch++;
4601 diff_free_filepair(p);
4602 }
4603 }
4604 free(q->queue);
4605 *q = outq;
4606}
4607
4608static int diffnamecmp(const void *a_, const void *b_)
4609{
4610 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4611 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4612 const char *name_a, *name_b;
4613
4614 name_a = a->one ? a->one->path : a->two->path;
4615 name_b = b->one ? b->one->path : b->two->path;
4616 return strcmp(name_a, name_b);
4617}
4618
4619void diffcore_fix_diff_index(struct diff_options *options)
4620{
4621 struct diff_queue_struct *q = &diff_queued_diff;
4622 qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4623}
4624
4625void diffcore_std(struct diff_options *options)
4626{
4627 if (options->skip_stat_unmatch)
4628 diffcore_skip_stat_unmatch(options);
4629 if (!options->found_follow) {
4630 /* See try_to_follow_renames() in tree-diff.c */
4631 if (options->break_opt != -1)
4632 diffcore_break(options->break_opt);
4633 if (options->detect_rename)
4634 diffcore_rename(options);
4635 if (options->break_opt != -1)
4636 diffcore_merge_broken();
4637 }
4638 if (options->pickaxe)
4639 diffcore_pickaxe(options);
4640 if (options->orderfile)
4641 diffcore_order(options->orderfile);
4642 if (!options->found_follow)
4643 /* See try_to_follow_renames() in tree-diff.c */
4644 diff_resolve_rename_copy();
4645 diffcore_apply_filter(options->filter);
4646
4647 if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4648 DIFF_OPT_SET(options, HAS_CHANGES);
4649 else
4650 DIFF_OPT_CLR(options, HAS_CHANGES);
4651
4652 options->found_follow = 0;
4653}
4654
4655int diff_result_code(struct diff_options *opt, int status)
4656{
4657 int result = 0;
4658
4659 diff_warn_rename_limit("diff.renamelimit",
4660 opt->needed_rename_limit,
4661 opt->degraded_cc_to_c);
4662 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4663 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4664 return status;
4665 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4666 DIFF_OPT_TST(opt, HAS_CHANGES))
4667 result |= 01;
4668 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4669 DIFF_OPT_TST(opt, CHECK_FAILED))
4670 result |= 02;
4671 return result;
4672}
4673
4674int diff_can_quit_early(struct diff_options *opt)
4675{
4676 return (DIFF_OPT_TST(opt, QUICK) &&
4677 !opt->filter &&
4678 DIFF_OPT_TST(opt, HAS_CHANGES));
4679}
4680
4681/*
4682 * Shall changes to this submodule be ignored?
4683 *
4684 * Submodule changes can be configured to be ignored separately for each path,
4685 * but that configuration can be overridden from the command line.
4686 */
4687static int is_submodule_ignored(const char *path, struct diff_options *options)
4688{
4689 int ignored = 0;
4690 unsigned orig_flags = options->flags;
4691 if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4692 set_diffopt_flags_from_submodule_config(options, path);
4693 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4694 ignored = 1;
4695 options->flags = orig_flags;
4696 return ignored;
4697}
4698
4699void diff_addremove(struct diff_options *options,
4700 int addremove, unsigned mode,
4701 const unsigned char *sha1,
4702 const char *concatpath, unsigned dirty_submodule)
4703{
4704 struct diff_filespec *one, *two;
4705
4706 if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
4707 return;
4708
4709 /* This may look odd, but it is a preparation for
4710 * feeding "there are unchanged files which should
4711 * not produce diffs, but when you are doing copy
4712 * detection you would need them, so here they are"
4713 * entries to the diff-core. They will be prefixed
4714 * with something like '=' or '*' (I haven't decided
4715 * which but should not make any difference).
4716 * Feeding the same new and old to diff_change()
4717 * also has the same effect.
4718 * Before the final output happens, they are pruned after
4719 * merged into rename/copy pairs as appropriate.
4720 */
4721 if (DIFF_OPT_TST(options, REVERSE_DIFF))
4722 addremove = (addremove == '+' ? '-' :
4723 addremove == '-' ? '+' : addremove);
4724
4725 if (options->prefix &&
4726 strncmp(concatpath, options->prefix, options->prefix_length))
4727 return;
4728
4729 one = alloc_filespec(concatpath);
4730 two = alloc_filespec(concatpath);
4731
4732 if (addremove != '+')
4733 fill_filespec(one, sha1, mode);
4734 if (addremove != '-') {
4735 fill_filespec(two, sha1, mode);
4736 two->dirty_submodule = dirty_submodule;
4737 }
4738
4739 diff_queue(&diff_queued_diff, one, two);
4740 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4741 DIFF_OPT_SET(options, HAS_CHANGES);
4742}
4743
4744void diff_change(struct diff_options *options,
4745 unsigned old_mode, unsigned new_mode,
4746 const unsigned char *old_sha1,
4747 const unsigned char *new_sha1,
4748 const char *concatpath,
4749 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
4750{
4751 struct diff_filespec *one, *two;
4752
4753 if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
4754 is_submodule_ignored(concatpath, options))
4755 return;
4756
4757 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
4758 unsigned tmp;
4759 const unsigned char *tmp_c;
4760 tmp = old_mode; old_mode = new_mode; new_mode = tmp;
4761 tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
4762 tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
4763 new_dirty_submodule = tmp;
4764 }
4765
4766 if (options->prefix &&
4767 strncmp(concatpath, options->prefix, options->prefix_length))
4768 return;
4769
4770 one = alloc_filespec(concatpath);
4771 two = alloc_filespec(concatpath);
4772 fill_filespec(one, old_sha1, old_mode);
4773 fill_filespec(two, new_sha1, new_mode);
4774 one->dirty_submodule = old_dirty_submodule;
4775 two->dirty_submodule = new_dirty_submodule;
4776
4777 diff_queue(&diff_queued_diff, one, two);
4778 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4779 DIFF_OPT_SET(options, HAS_CHANGES);
4780}
4781
4782struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
4783{
4784 struct diff_filepair *pair;
4785 struct diff_filespec *one, *two;
4786
4787 if (options->prefix &&
4788 strncmp(path, options->prefix, options->prefix_length))
4789 return NULL;
4790
4791 one = alloc_filespec(path);
4792 two = alloc_filespec(path);
4793 pair = diff_queue(&diff_queued_diff, one, two);
4794 pair->is_unmerged = 1;
4795 return pair;
4796}
4797
4798static char *run_textconv(const char *pgm, struct diff_filespec *spec,
4799 size_t *outsize)
4800{
4801 struct diff_tempfile *temp;
4802 const char *argv[3];
4803 const char **arg = argv;
4804 struct child_process child;
4805 struct strbuf buf = STRBUF_INIT;
4806 int err = 0;
4807
4808 temp = prepare_temp_file(spec->path, spec);
4809 *arg++ = pgm;
4810 *arg++ = temp->name;
4811 *arg = NULL;
4812
4813 memset(&child, 0, sizeof(child));
4814 child.use_shell = 1;
4815 child.argv = argv;
4816 child.out = -1;
4817 if (start_command(&child)) {
4818 remove_tempfile();
4819 return NULL;
4820 }
4821
4822 if (strbuf_read(&buf, child.out, 0) < 0)
4823 err = error("error reading from textconv command '%s'", pgm);
4824 close(child.out);
4825
4826 if (finish_command(&child) || err) {
4827 strbuf_release(&buf);
4828 remove_tempfile();
4829 return NULL;
4830 }
4831 remove_tempfile();
4832
4833 return strbuf_detach(&buf, outsize);
4834}
4835
4836size_t fill_textconv(struct userdiff_driver *driver,
4837 struct diff_filespec *df,
4838 char **outbuf)
4839{
4840 size_t size;
4841
4842 if (!driver || !driver->textconv) {
4843 if (!DIFF_FILE_VALID(df)) {
4844 *outbuf = "";
4845 return 0;
4846 }
4847 if (diff_populate_filespec(df, 0))
4848 die("unable to read files to diff");
4849 *outbuf = df->data;
4850 return df->size;
4851 }
4852
4853 if (driver->textconv_cache && df->sha1_valid) {
4854 *outbuf = notes_cache_get(driver->textconv_cache, df->sha1,
4855 &size);
4856 if (*outbuf)
4857 return size;
4858 }
4859
4860 *outbuf = run_textconv(driver->textconv, df, &size);
4861 if (!*outbuf)
4862 die("unable to read files to diff");
4863
4864 if (driver->textconv_cache && df->sha1_valid) {
4865 /* ignore errors, as we might be in a readonly repository */
4866 notes_cache_put(driver->textconv_cache, df->sha1, *outbuf,
4867 size);
4868 /*
4869 * we could save up changes and flush them all at the end,
4870 * but we would need an extra call after all diffing is done.
4871 * Since generating a cache entry is the slow path anyway,
4872 * this extra overhead probably isn't a big deal.
4873 */
4874 notes_cache_write(driver->textconv_cache);
4875 }
4876
4877 return size;
4878}