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