1d0f8b509ff400f28bd7717a98415b477835c942
1/*
2 * Copyright (C) 2005 Junio C Hamano
3 */
4#include "cache.h"
5#include "config.h"
6#include "tempfile.h"
7#include "quote.h"
8#include "diff.h"
9#include "diffcore.h"
10#include "delta.h"
11#include "xdiff-interface.h"
12#include "color.h"
13#include "attr.h"
14#include "run-command.h"
15#include "utf8.h"
16#include "userdiff.h"
17#include "submodule-config.h"
18#include "submodule.h"
19#include "ll-merge.h"
20#include "string-list.h"
21#include "argv-array.h"
22#include "graph.h"
23
24#ifdef NO_FAST_WORKING_DIRECTORY
25#define FAST_WORKING_DIRECTORY 0
26#else
27#define FAST_WORKING_DIRECTORY 1
28#endif
29
30static int diff_detect_rename_default;
31static int diff_indent_heuristic = 1;
32static int diff_rename_limit_default = 400;
33static int diff_suppress_blank_empty;
34static int diff_use_color_default = -1;
35static int diff_context_default = 3;
36static int diff_interhunk_context_default;
37static const char *diff_word_regex_cfg;
38static const char *external_diff_cmd_cfg;
39static const char *diff_order_file_cfg;
40int diff_auto_refresh_index = 1;
41static int diff_mnemonic_prefix;
42static int diff_no_prefix;
43static int diff_stat_graph_width;
44static int diff_dirstat_permille_default = 30;
45static struct diff_options default_diff_options;
46static long diff_algorithm;
47static unsigned ws_error_highlight_default = WSEH_NEW;
48
49static char diff_colors[][COLOR_MAXLEN] = {
50 GIT_COLOR_RESET,
51 GIT_COLOR_NORMAL, /* CONTEXT */
52 GIT_COLOR_BOLD, /* METAINFO */
53 GIT_COLOR_CYAN, /* FRAGINFO */
54 GIT_COLOR_RED, /* OLD */
55 GIT_COLOR_GREEN, /* NEW */
56 GIT_COLOR_YELLOW, /* COMMIT */
57 GIT_COLOR_BG_RED, /* WHITESPACE */
58 GIT_COLOR_NORMAL, /* FUNCINFO */
59};
60
61static NORETURN void die_want_option(const char *option_name)
62{
63 die(_("option '%s' requires a value"), option_name);
64}
65
66static int parse_diff_color_slot(const char *var)
67{
68 if (!strcasecmp(var, "context") || !strcasecmp(var, "plain"))
69 return DIFF_CONTEXT;
70 if (!strcasecmp(var, "meta"))
71 return DIFF_METAINFO;
72 if (!strcasecmp(var, "frag"))
73 return DIFF_FRAGINFO;
74 if (!strcasecmp(var, "old"))
75 return DIFF_FILE_OLD;
76 if (!strcasecmp(var, "new"))
77 return DIFF_FILE_NEW;
78 if (!strcasecmp(var, "commit"))
79 return DIFF_COMMIT;
80 if (!strcasecmp(var, "whitespace"))
81 return DIFF_WHITESPACE;
82 if (!strcasecmp(var, "func"))
83 return DIFF_FUNCINFO;
84 return -1;
85}
86
87static int parse_dirstat_params(struct diff_options *options, const char *params_string,
88 struct strbuf *errmsg)
89{
90 char *params_copy = xstrdup(params_string);
91 struct string_list params = STRING_LIST_INIT_NODUP;
92 int ret = 0;
93 int i;
94
95 if (*params_copy)
96 string_list_split_in_place(¶ms, params_copy, ',', -1);
97 for (i = 0; i < params.nr; i++) {
98 const char *p = params.items[i].string;
99 if (!strcmp(p, "changes")) {
100 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
101 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
102 } else if (!strcmp(p, "lines")) {
103 DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
104 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
105 } else if (!strcmp(p, "files")) {
106 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
107 DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
108 } else if (!strcmp(p, "noncumulative")) {
109 DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
110 } else if (!strcmp(p, "cumulative")) {
111 DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
112 } else if (isdigit(*p)) {
113 char *end;
114 int permille = strtoul(p, &end, 10) * 10;
115 if (*end == '.' && isdigit(*++end)) {
116 /* only use first digit */
117 permille += *end - '0';
118 /* .. and ignore any further digits */
119 while (isdigit(*++end))
120 ; /* nothing */
121 }
122 if (!*end)
123 options->dirstat_permille = permille;
124 else {
125 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%s'\n"),
126 p);
127 ret++;
128 }
129 } else {
130 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%s'\n"), p);
131 ret++;
132 }
133
134 }
135 string_list_clear(¶ms, 0);
136 free(params_copy);
137 return ret;
138}
139
140static int parse_submodule_params(struct diff_options *options, const char *value)
141{
142 if (!strcmp(value, "log"))
143 options->submodule_format = DIFF_SUBMODULE_LOG;
144 else if (!strcmp(value, "short"))
145 options->submodule_format = DIFF_SUBMODULE_SHORT;
146 else if (!strcmp(value, "diff"))
147 options->submodule_format = DIFF_SUBMODULE_INLINE_DIFF;
148 else
149 return -1;
150 return 0;
151}
152
153static int git_config_rename(const char *var, const char *value)
154{
155 if (!value)
156 return DIFF_DETECT_RENAME;
157 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
158 return DIFF_DETECT_COPY;
159 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
160}
161
162long parse_algorithm_value(const char *value)
163{
164 if (!value)
165 return -1;
166 else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
167 return 0;
168 else if (!strcasecmp(value, "minimal"))
169 return XDF_NEED_MINIMAL;
170 else if (!strcasecmp(value, "patience"))
171 return XDF_PATIENCE_DIFF;
172 else if (!strcasecmp(value, "histogram"))
173 return XDF_HISTOGRAM_DIFF;
174 return -1;
175}
176
177static int parse_one_token(const char **arg, const char *token)
178{
179 const char *rest;
180 if (skip_prefix(*arg, token, &rest) && (!*rest || *rest == ',')) {
181 *arg = rest;
182 return 1;
183 }
184 return 0;
185}
186
187static int parse_ws_error_highlight(const char *arg)
188{
189 const char *orig_arg = arg;
190 unsigned val = 0;
191
192 while (*arg) {
193 if (parse_one_token(&arg, "none"))
194 val = 0;
195 else if (parse_one_token(&arg, "default"))
196 val = WSEH_NEW;
197 else if (parse_one_token(&arg, "all"))
198 val = WSEH_NEW | WSEH_OLD | WSEH_CONTEXT;
199 else if (parse_one_token(&arg, "new"))
200 val |= WSEH_NEW;
201 else if (parse_one_token(&arg, "old"))
202 val |= WSEH_OLD;
203 else if (parse_one_token(&arg, "context"))
204 val |= WSEH_CONTEXT;
205 else {
206 return -1 - (int)(arg - orig_arg);
207 }
208 if (*arg)
209 arg++;
210 }
211 return val;
212}
213
214/*
215 * These are to give UI layer defaults.
216 * The core-level commands such as git-diff-files should
217 * never be affected by the setting of diff.renames
218 * the user happens to have in the configuration file.
219 */
220void init_diff_ui_defaults(void)
221{
222 diff_detect_rename_default = 1;
223}
224
225int git_diff_heuristic_config(const char *var, const char *value, void *cb)
226{
227 if (!strcmp(var, "diff.indentheuristic"))
228 diff_indent_heuristic = git_config_bool(var, value);
229 return 0;
230}
231
232int git_diff_ui_config(const char *var, const char *value, void *cb)
233{
234 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
235 diff_use_color_default = git_config_colorbool(var, value);
236 return 0;
237 }
238 if (!strcmp(var, "diff.context")) {
239 diff_context_default = git_config_int(var, value);
240 if (diff_context_default < 0)
241 return -1;
242 return 0;
243 }
244 if (!strcmp(var, "diff.interhunkcontext")) {
245 diff_interhunk_context_default = git_config_int(var, value);
246 if (diff_interhunk_context_default < 0)
247 return -1;
248 return 0;
249 }
250 if (!strcmp(var, "diff.renames")) {
251 diff_detect_rename_default = git_config_rename(var, value);
252 return 0;
253 }
254 if (!strcmp(var, "diff.autorefreshindex")) {
255 diff_auto_refresh_index = git_config_bool(var, value);
256 return 0;
257 }
258 if (!strcmp(var, "diff.mnemonicprefix")) {
259 diff_mnemonic_prefix = git_config_bool(var, value);
260 return 0;
261 }
262 if (!strcmp(var, "diff.noprefix")) {
263 diff_no_prefix = git_config_bool(var, value);
264 return 0;
265 }
266 if (!strcmp(var, "diff.statgraphwidth")) {
267 diff_stat_graph_width = git_config_int(var, value);
268 return 0;
269 }
270 if (!strcmp(var, "diff.external"))
271 return git_config_string(&external_diff_cmd_cfg, var, value);
272 if (!strcmp(var, "diff.wordregex"))
273 return git_config_string(&diff_word_regex_cfg, var, value);
274 if (!strcmp(var, "diff.orderfile"))
275 return git_config_pathname(&diff_order_file_cfg, var, value);
276
277 if (!strcmp(var, "diff.ignoresubmodules"))
278 handle_ignore_submodules_arg(&default_diff_options, value);
279
280 if (!strcmp(var, "diff.submodule")) {
281 if (parse_submodule_params(&default_diff_options, value))
282 warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
283 value);
284 return 0;
285 }
286
287 if (!strcmp(var, "diff.algorithm")) {
288 diff_algorithm = parse_algorithm_value(value);
289 if (diff_algorithm < 0)
290 return -1;
291 return 0;
292 }
293
294 if (!strcmp(var, "diff.wserrorhighlight")) {
295 int val = parse_ws_error_highlight(value);
296 if (val < 0)
297 return -1;
298 ws_error_highlight_default = val;
299 return 0;
300 }
301
302 if (git_color_config(var, value, cb) < 0)
303 return -1;
304
305 return git_diff_basic_config(var, value, cb);
306}
307
308int git_diff_basic_config(const char *var, const char *value, void *cb)
309{
310 const char *name;
311
312 if (!strcmp(var, "diff.renamelimit")) {
313 diff_rename_limit_default = git_config_int(var, value);
314 return 0;
315 }
316
317 if (userdiff_config(var, value) < 0)
318 return -1;
319
320 if (skip_prefix(var, "diff.color.", &name) ||
321 skip_prefix(var, "color.diff.", &name)) {
322 int slot = parse_diff_color_slot(name);
323 if (slot < 0)
324 return 0;
325 if (!value)
326 return config_error_nonbool(var);
327 return color_parse(value, diff_colors[slot]);
328 }
329
330 /* like GNU diff's --suppress-blank-empty option */
331 if (!strcmp(var, "diff.suppressblankempty") ||
332 /* for backwards compatibility */
333 !strcmp(var, "diff.suppress-blank-empty")) {
334 diff_suppress_blank_empty = git_config_bool(var, value);
335 return 0;
336 }
337
338 if (!strcmp(var, "diff.dirstat")) {
339 struct strbuf errmsg = STRBUF_INIT;
340 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
341 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
342 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
343 errmsg.buf);
344 strbuf_release(&errmsg);
345 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
346 return 0;
347 }
348
349 if (starts_with(var, "submodule."))
350 return parse_submodule_config_option(var, value);
351
352 if (git_diff_heuristic_config(var, value, cb) < 0)
353 return -1;
354
355 return git_default_config(var, value, cb);
356}
357
358static char *quote_two(const char *one, const char *two)
359{
360 int need_one = quote_c_style(one, NULL, NULL, 1);
361 int need_two = quote_c_style(two, NULL, NULL, 1);
362 struct strbuf res = STRBUF_INIT;
363
364 if (need_one + need_two) {
365 strbuf_addch(&res, '"');
366 quote_c_style(one, &res, NULL, 1);
367 quote_c_style(two, &res, NULL, 1);
368 strbuf_addch(&res, '"');
369 } else {
370 strbuf_addstr(&res, one);
371 strbuf_addstr(&res, two);
372 }
373 return strbuf_detach(&res, NULL);
374}
375
376static const char *external_diff(void)
377{
378 static const char *external_diff_cmd = NULL;
379 static int done_preparing = 0;
380
381 if (done_preparing)
382 return external_diff_cmd;
383 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
384 if (!external_diff_cmd)
385 external_diff_cmd = external_diff_cmd_cfg;
386 done_preparing = 1;
387 return external_diff_cmd;
388}
389
390/*
391 * Keep track of files used for diffing. Sometimes such an entry
392 * refers to a temporary file, sometimes to an existing file, and
393 * sometimes to "/dev/null".
394 */
395static struct diff_tempfile {
396 /*
397 * filename external diff should read from, or NULL if this
398 * entry is currently not in use:
399 */
400 const char *name;
401
402 char hex[GIT_MAX_HEXSZ + 1];
403 char mode[10];
404
405 /*
406 * If this diff_tempfile instance refers to a temporary file,
407 * this tempfile object is used to manage its lifetime.
408 */
409 struct tempfile tempfile;
410} diff_temp[2];
411
412typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
413
414struct emit_callback {
415 int color_diff;
416 unsigned ws_rule;
417 int blank_at_eof_in_preimage;
418 int blank_at_eof_in_postimage;
419 int lno_in_preimage;
420 int lno_in_postimage;
421 sane_truncate_fn truncate;
422 const char **label_path;
423 struct diff_words_data *diff_words;
424 struct diff_options *opt;
425 struct strbuf *header;
426};
427
428static int count_lines(const char *data, int size)
429{
430 int count, ch, completely_empty = 1, nl_just_seen = 0;
431 count = 0;
432 while (0 < size--) {
433 ch = *data++;
434 if (ch == '\n') {
435 count++;
436 nl_just_seen = 1;
437 completely_empty = 0;
438 }
439 else {
440 nl_just_seen = 0;
441 completely_empty = 0;
442 }
443 }
444 if (completely_empty)
445 return 0;
446 if (!nl_just_seen)
447 count++; /* no trailing newline */
448 return count;
449}
450
451static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
452{
453 if (!DIFF_FILE_VALID(one)) {
454 mf->ptr = (char *)""; /* does not matter */
455 mf->size = 0;
456 return 0;
457 }
458 else if (diff_populate_filespec(one, 0))
459 return -1;
460
461 mf->ptr = one->data;
462 mf->size = one->size;
463 return 0;
464}
465
466/* like fill_mmfile, but only for size, so we can avoid retrieving blob */
467static unsigned long diff_filespec_size(struct diff_filespec *one)
468{
469 if (!DIFF_FILE_VALID(one))
470 return 0;
471 diff_populate_filespec(one, CHECK_SIZE_ONLY);
472 return one->size;
473}
474
475static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
476{
477 char *ptr = mf->ptr;
478 long size = mf->size;
479 int cnt = 0;
480
481 if (!size)
482 return cnt;
483 ptr += size - 1; /* pointing at the very end */
484 if (*ptr != '\n')
485 ; /* incomplete line */
486 else
487 ptr--; /* skip the last LF */
488 while (mf->ptr < ptr) {
489 char *prev_eol;
490 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
491 if (*prev_eol == '\n')
492 break;
493 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
494 break;
495 cnt++;
496 ptr = prev_eol - 1;
497 }
498 return cnt;
499}
500
501static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
502 struct emit_callback *ecbdata)
503{
504 int l1, l2, at;
505 unsigned ws_rule = ecbdata->ws_rule;
506 l1 = count_trailing_blank(mf1, ws_rule);
507 l2 = count_trailing_blank(mf2, ws_rule);
508 if (l2 <= l1) {
509 ecbdata->blank_at_eof_in_preimage = 0;
510 ecbdata->blank_at_eof_in_postimage = 0;
511 return;
512 }
513 at = count_lines(mf1->ptr, mf1->size);
514 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
515
516 at = count_lines(mf2->ptr, mf2->size);
517 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
518}
519
520static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
521 int first, const char *line, int len)
522{
523 int has_trailing_newline, has_trailing_carriage_return;
524 int nofirst;
525 FILE *file = o->file;
526
527 fputs(diff_line_prefix(o), file);
528
529 if (len == 0) {
530 has_trailing_newline = (first == '\n');
531 has_trailing_carriage_return = (!has_trailing_newline &&
532 (first == '\r'));
533 nofirst = has_trailing_newline || has_trailing_carriage_return;
534 } else {
535 has_trailing_newline = (len > 0 && line[len-1] == '\n');
536 if (has_trailing_newline)
537 len--;
538 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
539 if (has_trailing_carriage_return)
540 len--;
541 nofirst = 0;
542 }
543
544 if (len || !nofirst) {
545 fputs(set, file);
546 if (!nofirst)
547 fputc(first, file);
548 fwrite(line, len, 1, file);
549 fputs(reset, file);
550 }
551 if (has_trailing_carriage_return)
552 fputc('\r', file);
553 if (has_trailing_newline)
554 fputc('\n', file);
555}
556
557static void emit_line(struct diff_options *o, const char *set, const char *reset,
558 const char *line, int len)
559{
560 emit_line_0(o, set, reset, line[0], line+1, len-1);
561}
562
563enum diff_symbol {
564 DIFF_SYMBOL_BINARY_FILES,
565 DIFF_SYMBOL_HEADER,
566 DIFF_SYMBOL_FILEPAIR_PLUS,
567 DIFF_SYMBOL_FILEPAIR_MINUS,
568 DIFF_SYMBOL_WORDS_PORCELAIN,
569 DIFF_SYMBOL_WORDS,
570 DIFF_SYMBOL_CONTEXT,
571 DIFF_SYMBOL_CONTEXT_INCOMPLETE,
572 DIFF_SYMBOL_PLUS,
573 DIFF_SYMBOL_MINUS,
574 DIFF_SYMBOL_NO_LF_EOF,
575 DIFF_SYMBOL_CONTEXT_FRAGINFO,
576 DIFF_SYMBOL_CONTEXT_MARKER,
577 DIFF_SYMBOL_SEPARATOR
578};
579/*
580 * Flags for content lines:
581 * 0..12 are whitespace rules
582 * 13-15 are WSEH_NEW | WSEH_OLD | WSEH_CONTEXT
583 * 16 is marking if the line is blank at EOF
584 */
585#define DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF (1<<16)
586#define DIFF_SYMBOL_CONTENT_WS_MASK (WSEH_NEW | WSEH_OLD | WSEH_CONTEXT | WS_RULE_MASK)
587
588static void emit_line_ws_markup(struct diff_options *o,
589 const char *set, const char *reset,
590 const char *line, int len, char sign,
591 unsigned ws_rule, int blank_at_eof)
592{
593 const char *ws = NULL;
594
595 if (o->ws_error_highlight & ws_rule) {
596 ws = diff_get_color_opt(o, DIFF_WHITESPACE);
597 if (!*ws)
598 ws = NULL;
599 }
600
601 if (!ws)
602 emit_line_0(o, set, reset, sign, line, len);
603 else if (blank_at_eof)
604 /* Blank line at EOF - paint '+' as well */
605 emit_line_0(o, ws, reset, sign, line, len);
606 else {
607 /* Emit just the prefix, then the rest. */
608 emit_line_0(o, set, reset, sign, "", 0);
609 ws_check_emit(line, len, ws_rule,
610 o->file, set, reset, ws);
611 }
612}
613
614static void emit_diff_symbol(struct diff_options *o, enum diff_symbol s,
615 const char *line, int len, unsigned flags)
616{
617 static const char *nneof = " No newline at end of file\n";
618 const char *context, *reset, *set, *meta;
619 switch (s) {
620 case DIFF_SYMBOL_NO_LF_EOF:
621 context = diff_get_color_opt(o, DIFF_CONTEXT);
622 reset = diff_get_color_opt(o, DIFF_RESET);
623 putc('\n', o->file);
624 emit_line_0(o, context, reset, '\\',
625 nneof, strlen(nneof));
626 break;
627 case DIFF_SYMBOL_CONTEXT_FRAGINFO:
628 emit_line(o, "", "", line, len);
629 break;
630 case DIFF_SYMBOL_CONTEXT_INCOMPLETE:
631 case DIFF_SYMBOL_CONTEXT_MARKER:
632 context = diff_get_color_opt(o, DIFF_CONTEXT);
633 reset = diff_get_color_opt(o, DIFF_RESET);
634 emit_line(o, context, reset, line, len);
635 break;
636 case DIFF_SYMBOL_SEPARATOR:
637 fprintf(o->file, "%s%c",
638 diff_line_prefix(o),
639 o->line_termination);
640 break;
641 case DIFF_SYMBOL_CONTEXT:
642 set = diff_get_color_opt(o, DIFF_CONTEXT);
643 reset = diff_get_color_opt(o, DIFF_RESET);
644 emit_line_ws_markup(o, set, reset, line, len, ' ',
645 flags & (DIFF_SYMBOL_CONTENT_WS_MASK), 0);
646 break;
647 case DIFF_SYMBOL_PLUS:
648 set = diff_get_color_opt(o, DIFF_FILE_NEW);
649 reset = diff_get_color_opt(o, DIFF_RESET);
650 emit_line_ws_markup(o, set, reset, line, len, '+',
651 flags & DIFF_SYMBOL_CONTENT_WS_MASK,
652 flags & DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF);
653 break;
654 case DIFF_SYMBOL_MINUS:
655 set = diff_get_color_opt(o, DIFF_FILE_OLD);
656 reset = diff_get_color_opt(o, DIFF_RESET);
657 emit_line_ws_markup(o, set, reset, line, len, '-',
658 flags & DIFF_SYMBOL_CONTENT_WS_MASK, 0);
659 break;
660 case DIFF_SYMBOL_WORDS_PORCELAIN:
661 context = diff_get_color_opt(o, DIFF_CONTEXT);
662 reset = diff_get_color_opt(o, DIFF_RESET);
663 emit_line(o, context, reset, line, len);
664 fputs("~\n", o->file);
665 break;
666 case DIFF_SYMBOL_WORDS:
667 context = diff_get_color_opt(o, DIFF_CONTEXT);
668 reset = diff_get_color_opt(o, DIFF_RESET);
669 /*
670 * Skip the prefix character, if any. With
671 * diff_suppress_blank_empty, there may be
672 * none.
673 */
674 if (line[0] != '\n') {
675 line++;
676 len--;
677 }
678 emit_line(o, context, reset, line, len);
679 break;
680 case DIFF_SYMBOL_FILEPAIR_PLUS:
681 meta = diff_get_color_opt(o, DIFF_METAINFO);
682 reset = diff_get_color_opt(o, DIFF_RESET);
683 fprintf(o->file, "%s%s+++ %s%s%s\n", diff_line_prefix(o), meta,
684 line, reset,
685 strchr(line, ' ') ? "\t" : "");
686 break;
687 case DIFF_SYMBOL_FILEPAIR_MINUS:
688 meta = diff_get_color_opt(o, DIFF_METAINFO);
689 reset = diff_get_color_opt(o, DIFF_RESET);
690 fprintf(o->file, "%s%s--- %s%s%s\n", diff_line_prefix(o), meta,
691 line, reset,
692 strchr(line, ' ') ? "\t" : "");
693 break;
694 case DIFF_SYMBOL_BINARY_FILES:
695 case DIFF_SYMBOL_HEADER:
696 fprintf(o->file, "%s", line);
697 break;
698 default:
699 die("BUG: unknown diff symbol");
700 }
701}
702
703static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
704{
705 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
706 ecbdata->blank_at_eof_in_preimage &&
707 ecbdata->blank_at_eof_in_postimage &&
708 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
709 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
710 return 0;
711 return ws_blank_line(line, len, ecbdata->ws_rule);
712}
713
714static void emit_add_line(const char *reset,
715 struct emit_callback *ecbdata,
716 const char *line, int len)
717{
718 unsigned flags = WSEH_NEW | ecbdata->ws_rule;
719 if (new_blank_line_at_eof(ecbdata, line, len))
720 flags |= DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF;
721
722 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_PLUS, line, len, flags);
723}
724
725static void emit_del_line(const char *reset,
726 struct emit_callback *ecbdata,
727 const char *line, int len)
728{
729 unsigned flags = WSEH_OLD | ecbdata->ws_rule;
730 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_MINUS, line, len, flags);
731}
732
733static void emit_context_line(const char *reset,
734 struct emit_callback *ecbdata,
735 const char *line, int len)
736{
737 unsigned flags = WSEH_CONTEXT | ecbdata->ws_rule;
738 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_CONTEXT, line, len, flags);
739}
740
741static void emit_hunk_header(struct emit_callback *ecbdata,
742 const char *line, int len)
743{
744 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
745 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
746 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
747 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
748 static const char atat[2] = { '@', '@' };
749 const char *cp, *ep;
750 struct strbuf msgbuf = STRBUF_INIT;
751 int org_len = len;
752 int i = 1;
753
754 /*
755 * As a hunk header must begin with "@@ -<old>, +<new> @@",
756 * it always is at least 10 bytes long.
757 */
758 if (len < 10 ||
759 memcmp(line, atat, 2) ||
760 !(ep = memmem(line + 2, len - 2, atat, 2))) {
761 emit_diff_symbol(ecbdata->opt,
762 DIFF_SYMBOL_CONTEXT_MARKER, line, len, 0);
763 return;
764 }
765 ep += 2; /* skip over @@ */
766
767 /* The hunk header in fraginfo color */
768 strbuf_addstr(&msgbuf, frag);
769 strbuf_add(&msgbuf, line, ep - line);
770 strbuf_addstr(&msgbuf, reset);
771
772 /*
773 * trailing "\r\n"
774 */
775 for ( ; i < 3; i++)
776 if (line[len - i] == '\r' || line[len - i] == '\n')
777 len--;
778
779 /* blank before the func header */
780 for (cp = ep; ep - line < len; ep++)
781 if (*ep != ' ' && *ep != '\t')
782 break;
783 if (ep != cp) {
784 strbuf_addstr(&msgbuf, context);
785 strbuf_add(&msgbuf, cp, ep - cp);
786 strbuf_addstr(&msgbuf, reset);
787 }
788
789 if (ep < line + len) {
790 strbuf_addstr(&msgbuf, func);
791 strbuf_add(&msgbuf, ep, line + len - ep);
792 strbuf_addstr(&msgbuf, reset);
793 }
794
795 strbuf_add(&msgbuf, line + len, org_len - len);
796 strbuf_complete_line(&msgbuf);
797 emit_diff_symbol(ecbdata->opt,
798 DIFF_SYMBOL_CONTEXT_FRAGINFO, msgbuf.buf, msgbuf.len, 0);
799 strbuf_release(&msgbuf);
800}
801
802static struct diff_tempfile *claim_diff_tempfile(void) {
803 int i;
804 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
805 if (!diff_temp[i].name)
806 return diff_temp + i;
807 die("BUG: diff is failing to clean up its tempfiles");
808}
809
810static void remove_tempfile(void)
811{
812 int i;
813 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
814 if (is_tempfile_active(&diff_temp[i].tempfile))
815 delete_tempfile(&diff_temp[i].tempfile);
816 diff_temp[i].name = NULL;
817 }
818}
819
820static void print_line_count(FILE *file, int count)
821{
822 switch (count) {
823 case 0:
824 fprintf(file, "0,0");
825 break;
826 case 1:
827 fprintf(file, "1");
828 break;
829 default:
830 fprintf(file, "1,%d", count);
831 break;
832 }
833}
834
835static void emit_rewrite_lines(struct emit_callback *ecb,
836 int prefix, const char *data, int size)
837{
838 const char *endp = NULL;
839 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
840
841 while (0 < size) {
842 int len;
843
844 endp = memchr(data, '\n', size);
845 len = endp ? (endp - data + 1) : size;
846 if (prefix != '+') {
847 ecb->lno_in_preimage++;
848 emit_del_line(reset, ecb, data, len);
849 } else {
850 ecb->lno_in_postimage++;
851 emit_add_line(reset, ecb, data, len);
852 }
853 size -= len;
854 data += len;
855 }
856 if (!endp)
857 emit_diff_symbol(ecb->opt, DIFF_SYMBOL_NO_LF_EOF, NULL, 0, 0);
858}
859
860static void emit_rewrite_diff(const char *name_a,
861 const char *name_b,
862 struct diff_filespec *one,
863 struct diff_filespec *two,
864 struct userdiff_driver *textconv_one,
865 struct userdiff_driver *textconv_two,
866 struct diff_options *o)
867{
868 int lc_a, lc_b;
869 const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
870 const char *reset = diff_get_color(o->use_color, DIFF_RESET);
871 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
872 const char *a_prefix, *b_prefix;
873 char *data_one, *data_two;
874 size_t size_one, size_two;
875 struct emit_callback ecbdata;
876 const char *line_prefix = diff_line_prefix(o);
877
878 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
879 a_prefix = o->b_prefix;
880 b_prefix = o->a_prefix;
881 } else {
882 a_prefix = o->a_prefix;
883 b_prefix = o->b_prefix;
884 }
885
886 name_a += (*name_a == '/');
887 name_b += (*name_b == '/');
888
889 strbuf_reset(&a_name);
890 strbuf_reset(&b_name);
891 quote_two_c_style(&a_name, a_prefix, name_a, 0);
892 quote_two_c_style(&b_name, b_prefix, name_b, 0);
893
894 size_one = fill_textconv(textconv_one, one, &data_one);
895 size_two = fill_textconv(textconv_two, two, &data_two);
896
897 memset(&ecbdata, 0, sizeof(ecbdata));
898 ecbdata.color_diff = want_color(o->use_color);
899 ecbdata.ws_rule = whitespace_rule(name_b);
900 ecbdata.opt = o;
901 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
902 mmfile_t mf1, mf2;
903 mf1.ptr = (char *)data_one;
904 mf2.ptr = (char *)data_two;
905 mf1.size = size_one;
906 mf2.size = size_two;
907 check_blank_at_eof(&mf1, &mf2, &ecbdata);
908 }
909 ecbdata.lno_in_preimage = 1;
910 ecbdata.lno_in_postimage = 1;
911
912 lc_a = count_lines(data_one, size_one);
913 lc_b = count_lines(data_two, size_two);
914
915 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
916 a_name.buf, a_name.len, 0);
917 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
918 b_name.buf, b_name.len, 0);
919
920 fprintf(o->file, "%s%s@@ -", line_prefix, fraginfo);
921 if (!o->irreversible_delete)
922 print_line_count(o->file, lc_a);
923 else
924 fprintf(o->file, "?,?");
925 fprintf(o->file, " +");
926 print_line_count(o->file, lc_b);
927 fprintf(o->file, " @@%s\n", reset);
928 if (lc_a && !o->irreversible_delete)
929 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
930 if (lc_b)
931 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
932 if (textconv_one)
933 free((char *)data_one);
934 if (textconv_two)
935 free((char *)data_two);
936}
937
938struct diff_words_buffer {
939 mmfile_t text;
940 long alloc;
941 struct diff_words_orig {
942 const char *begin, *end;
943 } *orig;
944 int orig_nr, orig_alloc;
945};
946
947static void diff_words_append(char *line, unsigned long len,
948 struct diff_words_buffer *buffer)
949{
950 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
951 line++;
952 len--;
953 memcpy(buffer->text.ptr + buffer->text.size, line, len);
954 buffer->text.size += len;
955 buffer->text.ptr[buffer->text.size] = '\0';
956}
957
958struct diff_words_style_elem {
959 const char *prefix;
960 const char *suffix;
961 const char *color; /* NULL; filled in by the setup code if
962 * color is enabled */
963};
964
965struct diff_words_style {
966 enum diff_words_type type;
967 struct diff_words_style_elem new, old, ctx;
968 const char *newline;
969};
970
971static struct diff_words_style diff_words_styles[] = {
972 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
973 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
974 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
975};
976
977struct diff_words_data {
978 struct diff_words_buffer minus, plus;
979 const char *current_plus;
980 int last_minus;
981 struct diff_options *opt;
982 regex_t *word_regex;
983 enum diff_words_type type;
984 struct diff_words_style *style;
985};
986
987static int fn_out_diff_words_write_helper(FILE *fp,
988 struct diff_words_style_elem *st_el,
989 const char *newline,
990 size_t count, const char *buf,
991 const char *line_prefix)
992{
993 int print = 0;
994
995 while (count) {
996 char *p = memchr(buf, '\n', count);
997 if (print)
998 fputs(line_prefix, fp);
999 if (p != buf) {
1000 if (st_el->color && fputs(st_el->color, fp) < 0)
1001 return -1;
1002 if (fputs(st_el->prefix, fp) < 0 ||
1003 fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
1004 fputs(st_el->suffix, fp) < 0)
1005 return -1;
1006 if (st_el->color && *st_el->color
1007 && fputs(GIT_COLOR_RESET, fp) < 0)
1008 return -1;
1009 }
1010 if (!p)
1011 return 0;
1012 if (fputs(newline, fp) < 0)
1013 return -1;
1014 count -= p + 1 - buf;
1015 buf = p + 1;
1016 print = 1;
1017 }
1018 return 0;
1019}
1020
1021/*
1022 * '--color-words' algorithm can be described as:
1023 *
1024 * 1. collect the minus/plus lines of a diff hunk, divided into
1025 * minus-lines and plus-lines;
1026 *
1027 * 2. break both minus-lines and plus-lines into words and
1028 * place them into two mmfile_t with one word for each line;
1029 *
1030 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
1031 *
1032 * And for the common parts of the both file, we output the plus side text.
1033 * diff_words->current_plus is used to trace the current position of the plus file
1034 * which printed. diff_words->last_minus is used to trace the last minus word
1035 * printed.
1036 *
1037 * For '--graph' to work with '--color-words', we need to output the graph prefix
1038 * on each line of color words output. Generally, there are two conditions on
1039 * which we should output the prefix.
1040 *
1041 * 1. diff_words->last_minus == 0 &&
1042 * diff_words->current_plus == diff_words->plus.text.ptr
1043 *
1044 * that is: the plus text must start as a new line, and if there is no minus
1045 * word printed, a graph prefix must be printed.
1046 *
1047 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
1048 * *(diff_words->current_plus - 1) == '\n'
1049 *
1050 * that is: a graph prefix must be printed following a '\n'
1051 */
1052static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
1053{
1054 if ((diff_words->last_minus == 0 &&
1055 diff_words->current_plus == diff_words->plus.text.ptr) ||
1056 (diff_words->current_plus > diff_words->plus.text.ptr &&
1057 *(diff_words->current_plus - 1) == '\n')) {
1058 return 1;
1059 } else {
1060 return 0;
1061 }
1062}
1063
1064static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
1065{
1066 struct diff_words_data *diff_words = priv;
1067 struct diff_words_style *style = diff_words->style;
1068 int minus_first, minus_len, plus_first, plus_len;
1069 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
1070 struct diff_options *opt = diff_words->opt;
1071 const char *line_prefix;
1072
1073 if (line[0] != '@' || parse_hunk_header(line, len,
1074 &minus_first, &minus_len, &plus_first, &plus_len))
1075 return;
1076
1077 assert(opt);
1078 line_prefix = diff_line_prefix(opt);
1079
1080 /* POSIX requires that first be decremented by one if len == 0... */
1081 if (minus_len) {
1082 minus_begin = diff_words->minus.orig[minus_first].begin;
1083 minus_end =
1084 diff_words->minus.orig[minus_first + minus_len - 1].end;
1085 } else
1086 minus_begin = minus_end =
1087 diff_words->minus.orig[minus_first].end;
1088
1089 if (plus_len) {
1090 plus_begin = diff_words->plus.orig[plus_first].begin;
1091 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
1092 } else
1093 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
1094
1095 if (color_words_output_graph_prefix(diff_words)) {
1096 fputs(line_prefix, diff_words->opt->file);
1097 }
1098 if (diff_words->current_plus != plus_begin) {
1099 fn_out_diff_words_write_helper(diff_words->opt->file,
1100 &style->ctx, style->newline,
1101 plus_begin - diff_words->current_plus,
1102 diff_words->current_plus, line_prefix);
1103 if (*(plus_begin - 1) == '\n')
1104 fputs(line_prefix, diff_words->opt->file);
1105 }
1106 if (minus_begin != minus_end) {
1107 fn_out_diff_words_write_helper(diff_words->opt->file,
1108 &style->old, style->newline,
1109 minus_end - minus_begin, minus_begin,
1110 line_prefix);
1111 }
1112 if (plus_begin != plus_end) {
1113 fn_out_diff_words_write_helper(diff_words->opt->file,
1114 &style->new, style->newline,
1115 plus_end - plus_begin, plus_begin,
1116 line_prefix);
1117 }
1118
1119 diff_words->current_plus = plus_end;
1120 diff_words->last_minus = minus_first;
1121}
1122
1123/* This function starts looking at *begin, and returns 0 iff a word was found. */
1124static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
1125 int *begin, int *end)
1126{
1127 if (word_regex && *begin < buffer->size) {
1128 regmatch_t match[1];
1129 if (!regexec_buf(word_regex, buffer->ptr + *begin,
1130 buffer->size - *begin, 1, match, 0)) {
1131 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
1132 '\n', match[0].rm_eo - match[0].rm_so);
1133 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
1134 *begin += match[0].rm_so;
1135 return *begin >= *end;
1136 }
1137 return -1;
1138 }
1139
1140 /* find the next word */
1141 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
1142 (*begin)++;
1143 if (*begin >= buffer->size)
1144 return -1;
1145
1146 /* find the end of the word */
1147 *end = *begin + 1;
1148 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
1149 (*end)++;
1150
1151 return 0;
1152}
1153
1154/*
1155 * This function splits the words in buffer->text, stores the list with
1156 * newline separator into out, and saves the offsets of the original words
1157 * in buffer->orig.
1158 */
1159static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
1160 regex_t *word_regex)
1161{
1162 int i, j;
1163 long alloc = 0;
1164
1165 out->size = 0;
1166 out->ptr = NULL;
1167
1168 /* fake an empty "0th" word */
1169 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
1170 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
1171 buffer->orig_nr = 1;
1172
1173 for (i = 0; i < buffer->text.size; i++) {
1174 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
1175 return;
1176
1177 /* store original boundaries */
1178 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
1179 buffer->orig_alloc);
1180 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
1181 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
1182 buffer->orig_nr++;
1183
1184 /* store one word */
1185 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
1186 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
1187 out->ptr[out->size + j - i] = '\n';
1188 out->size += j - i + 1;
1189
1190 i = j - 1;
1191 }
1192}
1193
1194/* this executes the word diff on the accumulated buffers */
1195static void diff_words_show(struct diff_words_data *diff_words)
1196{
1197 xpparam_t xpp;
1198 xdemitconf_t xecfg;
1199 mmfile_t minus, plus;
1200 struct diff_words_style *style = diff_words->style;
1201
1202 struct diff_options *opt = diff_words->opt;
1203 const char *line_prefix;
1204
1205 assert(opt);
1206 line_prefix = diff_line_prefix(opt);
1207
1208 /* special case: only removal */
1209 if (!diff_words->plus.text.size) {
1210 fputs(line_prefix, diff_words->opt->file);
1211 fn_out_diff_words_write_helper(diff_words->opt->file,
1212 &style->old, style->newline,
1213 diff_words->minus.text.size,
1214 diff_words->minus.text.ptr, line_prefix);
1215 diff_words->minus.text.size = 0;
1216 return;
1217 }
1218
1219 diff_words->current_plus = diff_words->plus.text.ptr;
1220 diff_words->last_minus = 0;
1221
1222 memset(&xpp, 0, sizeof(xpp));
1223 memset(&xecfg, 0, sizeof(xecfg));
1224 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1225 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1226 xpp.flags = 0;
1227 /* as only the hunk header will be parsed, we need a 0-context */
1228 xecfg.ctxlen = 0;
1229 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1230 &xpp, &xecfg))
1231 die("unable to generate word diff");
1232 free(minus.ptr);
1233 free(plus.ptr);
1234 if (diff_words->current_plus != diff_words->plus.text.ptr +
1235 diff_words->plus.text.size) {
1236 if (color_words_output_graph_prefix(diff_words))
1237 fputs(line_prefix, diff_words->opt->file);
1238 fn_out_diff_words_write_helper(diff_words->opt->file,
1239 &style->ctx, style->newline,
1240 diff_words->plus.text.ptr + diff_words->plus.text.size
1241 - diff_words->current_plus, diff_words->current_plus,
1242 line_prefix);
1243 }
1244 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1245}
1246
1247/* In "color-words" mode, show word-diff of words accumulated in the buffer */
1248static void diff_words_flush(struct emit_callback *ecbdata)
1249{
1250 if (ecbdata->diff_words->minus.text.size ||
1251 ecbdata->diff_words->plus.text.size)
1252 diff_words_show(ecbdata->diff_words);
1253}
1254
1255static void diff_filespec_load_driver(struct diff_filespec *one)
1256{
1257 /* Use already-loaded driver */
1258 if (one->driver)
1259 return;
1260
1261 if (S_ISREG(one->mode))
1262 one->driver = userdiff_find_by_path(one->path);
1263
1264 /* Fallback to default settings */
1265 if (!one->driver)
1266 one->driver = userdiff_find_by_name("default");
1267}
1268
1269static const char *userdiff_word_regex(struct diff_filespec *one)
1270{
1271 diff_filespec_load_driver(one);
1272 return one->driver->word_regex;
1273}
1274
1275static void init_diff_words_data(struct emit_callback *ecbdata,
1276 struct diff_options *orig_opts,
1277 struct diff_filespec *one,
1278 struct diff_filespec *two)
1279{
1280 int i;
1281 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1282 memcpy(o, orig_opts, sizeof(struct diff_options));
1283
1284 ecbdata->diff_words =
1285 xcalloc(1, sizeof(struct diff_words_data));
1286 ecbdata->diff_words->type = o->word_diff;
1287 ecbdata->diff_words->opt = o;
1288 if (!o->word_regex)
1289 o->word_regex = userdiff_word_regex(one);
1290 if (!o->word_regex)
1291 o->word_regex = userdiff_word_regex(two);
1292 if (!o->word_regex)
1293 o->word_regex = diff_word_regex_cfg;
1294 if (o->word_regex) {
1295 ecbdata->diff_words->word_regex = (regex_t *)
1296 xmalloc(sizeof(regex_t));
1297 if (regcomp(ecbdata->diff_words->word_regex,
1298 o->word_regex,
1299 REG_EXTENDED | REG_NEWLINE))
1300 die ("Invalid regular expression: %s",
1301 o->word_regex);
1302 }
1303 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1304 if (o->word_diff == diff_words_styles[i].type) {
1305 ecbdata->diff_words->style =
1306 &diff_words_styles[i];
1307 break;
1308 }
1309 }
1310 if (want_color(o->use_color)) {
1311 struct diff_words_style *st = ecbdata->diff_words->style;
1312 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1313 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1314 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
1315 }
1316}
1317
1318static void free_diff_words_data(struct emit_callback *ecbdata)
1319{
1320 if (ecbdata->diff_words) {
1321 diff_words_flush(ecbdata);
1322 free (ecbdata->diff_words->opt);
1323 free (ecbdata->diff_words->minus.text.ptr);
1324 free (ecbdata->diff_words->minus.orig);
1325 free (ecbdata->diff_words->plus.text.ptr);
1326 free (ecbdata->diff_words->plus.orig);
1327 if (ecbdata->diff_words->word_regex) {
1328 regfree(ecbdata->diff_words->word_regex);
1329 free(ecbdata->diff_words->word_regex);
1330 }
1331 FREE_AND_NULL(ecbdata->diff_words);
1332 }
1333}
1334
1335const char *diff_get_color(int diff_use_color, enum color_diff ix)
1336{
1337 if (want_color(diff_use_color))
1338 return diff_colors[ix];
1339 return "";
1340}
1341
1342const char *diff_line_prefix(struct diff_options *opt)
1343{
1344 struct strbuf *msgbuf;
1345 if (!opt->output_prefix)
1346 return "";
1347
1348 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1349 return msgbuf->buf;
1350}
1351
1352static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1353{
1354 const char *cp;
1355 unsigned long allot;
1356 size_t l = len;
1357
1358 if (ecb->truncate)
1359 return ecb->truncate(line, len);
1360 cp = line;
1361 allot = l;
1362 while (0 < l) {
1363 (void) utf8_width(&cp, &l);
1364 if (!cp)
1365 break; /* truncated in the middle? */
1366 }
1367 return allot - l;
1368}
1369
1370static void find_lno(const char *line, struct emit_callback *ecbdata)
1371{
1372 const char *p;
1373 ecbdata->lno_in_preimage = 0;
1374 ecbdata->lno_in_postimage = 0;
1375 p = strchr(line, '-');
1376 if (!p)
1377 return; /* cannot happen */
1378 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1379 p = strchr(p, '+');
1380 if (!p)
1381 return; /* cannot happen */
1382 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1383}
1384
1385static void fn_out_consume(void *priv, char *line, unsigned long len)
1386{
1387 struct emit_callback *ecbdata = priv;
1388 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1389 struct diff_options *o = ecbdata->opt;
1390
1391 o->found_changes = 1;
1392
1393 if (ecbdata->header) {
1394 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
1395 ecbdata->header->buf, ecbdata->header->len, 0);
1396 strbuf_reset(ecbdata->header);
1397 ecbdata->header = NULL;
1398 }
1399
1400 if (ecbdata->label_path[0]) {
1401 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
1402 ecbdata->label_path[0],
1403 strlen(ecbdata->label_path[0]), 0);
1404 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
1405 ecbdata->label_path[1],
1406 strlen(ecbdata->label_path[1]), 0);
1407 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1408 }
1409
1410 if (diff_suppress_blank_empty
1411 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1412 line[0] = '\n';
1413 len = 1;
1414 }
1415
1416 if (line[0] == '@') {
1417 if (ecbdata->diff_words)
1418 diff_words_flush(ecbdata);
1419 len = sane_truncate_line(ecbdata, line, len);
1420 find_lno(line, ecbdata);
1421 emit_hunk_header(ecbdata, line, len);
1422 return;
1423 }
1424
1425 if (ecbdata->diff_words) {
1426 enum diff_symbol s =
1427 ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN ?
1428 DIFF_SYMBOL_WORDS_PORCELAIN : DIFF_SYMBOL_WORDS;
1429 if (line[0] == '-') {
1430 diff_words_append(line, len,
1431 &ecbdata->diff_words->minus);
1432 return;
1433 } else if (line[0] == '+') {
1434 diff_words_append(line, len,
1435 &ecbdata->diff_words->plus);
1436 return;
1437 } else if (starts_with(line, "\\ ")) {
1438 /*
1439 * Eat the "no newline at eof" marker as if we
1440 * saw a "+" or "-" line with nothing on it,
1441 * and return without diff_words_flush() to
1442 * defer processing. If this is the end of
1443 * preimage, more "+" lines may come after it.
1444 */
1445 return;
1446 }
1447 diff_words_flush(ecbdata);
1448 emit_diff_symbol(o, s, line, len, 0);
1449 return;
1450 }
1451
1452 switch (line[0]) {
1453 case '+':
1454 ecbdata->lno_in_postimage++;
1455 emit_add_line(reset, ecbdata, line + 1, len - 1);
1456 break;
1457 case '-':
1458 ecbdata->lno_in_preimage++;
1459 emit_del_line(reset, ecbdata, line + 1, len - 1);
1460 break;
1461 case ' ':
1462 ecbdata->lno_in_postimage++;
1463 ecbdata->lno_in_preimage++;
1464 emit_context_line(reset, ecbdata, line + 1, len - 1);
1465 break;
1466 default:
1467 /* incomplete line at the end */
1468 ecbdata->lno_in_preimage++;
1469 emit_diff_symbol(o, DIFF_SYMBOL_CONTEXT_INCOMPLETE,
1470 line, len, 0);
1471 break;
1472 }
1473}
1474
1475static char *pprint_rename(const char *a, const char *b)
1476{
1477 const char *old = a;
1478 const char *new = b;
1479 struct strbuf name = STRBUF_INIT;
1480 int pfx_length, sfx_length;
1481 int pfx_adjust_for_slash;
1482 int len_a = strlen(a);
1483 int len_b = strlen(b);
1484 int a_midlen, b_midlen;
1485 int qlen_a = quote_c_style(a, NULL, NULL, 0);
1486 int qlen_b = quote_c_style(b, NULL, NULL, 0);
1487
1488 if (qlen_a || qlen_b) {
1489 quote_c_style(a, &name, NULL, 0);
1490 strbuf_addstr(&name, " => ");
1491 quote_c_style(b, &name, NULL, 0);
1492 return strbuf_detach(&name, NULL);
1493 }
1494
1495 /* Find common prefix */
1496 pfx_length = 0;
1497 while (*old && *new && *old == *new) {
1498 if (*old == '/')
1499 pfx_length = old - a + 1;
1500 old++;
1501 new++;
1502 }
1503
1504 /* Find common suffix */
1505 old = a + len_a;
1506 new = b + len_b;
1507 sfx_length = 0;
1508 /*
1509 * If there is a common prefix, it must end in a slash. In
1510 * that case we let this loop run 1 into the prefix to see the
1511 * same slash.
1512 *
1513 * If there is no common prefix, we cannot do this as it would
1514 * underrun the input strings.
1515 */
1516 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
1517 while (a + pfx_length - pfx_adjust_for_slash <= old &&
1518 b + pfx_length - pfx_adjust_for_slash <= new &&
1519 *old == *new) {
1520 if (*old == '/')
1521 sfx_length = len_a - (old - a);
1522 old--;
1523 new--;
1524 }
1525
1526 /*
1527 * pfx{mid-a => mid-b}sfx
1528 * {pfx-a => pfx-b}sfx
1529 * pfx{sfx-a => sfx-b}
1530 * name-a => name-b
1531 */
1532 a_midlen = len_a - pfx_length - sfx_length;
1533 b_midlen = len_b - pfx_length - sfx_length;
1534 if (a_midlen < 0)
1535 a_midlen = 0;
1536 if (b_midlen < 0)
1537 b_midlen = 0;
1538
1539 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1540 if (pfx_length + sfx_length) {
1541 strbuf_add(&name, a, pfx_length);
1542 strbuf_addch(&name, '{');
1543 }
1544 strbuf_add(&name, a + pfx_length, a_midlen);
1545 strbuf_addstr(&name, " => ");
1546 strbuf_add(&name, b + pfx_length, b_midlen);
1547 if (pfx_length + sfx_length) {
1548 strbuf_addch(&name, '}');
1549 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1550 }
1551 return strbuf_detach(&name, NULL);
1552}
1553
1554struct diffstat_t {
1555 int nr;
1556 int alloc;
1557 struct diffstat_file {
1558 char *from_name;
1559 char *name;
1560 char *print_name;
1561 unsigned is_unmerged:1;
1562 unsigned is_binary:1;
1563 unsigned is_renamed:1;
1564 unsigned is_interesting:1;
1565 uintmax_t added, deleted;
1566 } **files;
1567};
1568
1569static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1570 const char *name_a,
1571 const char *name_b)
1572{
1573 struct diffstat_file *x;
1574 x = xcalloc(1, sizeof(*x));
1575 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
1576 diffstat->files[diffstat->nr++] = x;
1577 if (name_b) {
1578 x->from_name = xstrdup(name_a);
1579 x->name = xstrdup(name_b);
1580 x->is_renamed = 1;
1581 }
1582 else {
1583 x->from_name = NULL;
1584 x->name = xstrdup(name_a);
1585 }
1586 return x;
1587}
1588
1589static void diffstat_consume(void *priv, char *line, unsigned long len)
1590{
1591 struct diffstat_t *diffstat = priv;
1592 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1593
1594 if (line[0] == '+')
1595 x->added++;
1596 else if (line[0] == '-')
1597 x->deleted++;
1598}
1599
1600const char mime_boundary_leader[] = "------------";
1601
1602static int scale_linear(int it, int width, int max_change)
1603{
1604 if (!it)
1605 return 0;
1606 /*
1607 * make sure that at least one '-' or '+' is printed if
1608 * there is any change to this path. The easiest way is to
1609 * scale linearly as if the alloted width is one column shorter
1610 * than it is, and then add 1 to the result.
1611 */
1612 return 1 + (it * (width - 1) / max_change);
1613}
1614
1615static void show_name(FILE *file,
1616 const char *prefix, const char *name, int len)
1617{
1618 fprintf(file, " %s%-*s |", prefix, len, name);
1619}
1620
1621static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1622{
1623 if (cnt <= 0)
1624 return;
1625 fprintf(file, "%s", set);
1626 while (cnt--)
1627 putc(ch, file);
1628 fprintf(file, "%s", reset);
1629}
1630
1631static void fill_print_name(struct diffstat_file *file)
1632{
1633 char *pname;
1634
1635 if (file->print_name)
1636 return;
1637
1638 if (!file->is_renamed) {
1639 struct strbuf buf = STRBUF_INIT;
1640 if (quote_c_style(file->name, &buf, NULL, 0)) {
1641 pname = strbuf_detach(&buf, NULL);
1642 } else {
1643 pname = file->name;
1644 strbuf_release(&buf);
1645 }
1646 } else {
1647 pname = pprint_rename(file->from_name, file->name);
1648 }
1649 file->print_name = pname;
1650}
1651
1652int print_stat_summary(FILE *fp, int files, int insertions, int deletions)
1653{
1654 struct strbuf sb = STRBUF_INIT;
1655 int ret;
1656
1657 if (!files) {
1658 assert(insertions == 0 && deletions == 0);
1659 return fprintf(fp, "%s\n", " 0 files changed");
1660 }
1661
1662 strbuf_addf(&sb,
1663 (files == 1) ? " %d file changed" : " %d files changed",
1664 files);
1665
1666 /*
1667 * For binary diff, the caller may want to print "x files
1668 * changed" with insertions == 0 && deletions == 0.
1669 *
1670 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
1671 * is probably less confusing (i.e skip over "2 files changed
1672 * but nothing about added/removed lines? Is this a bug in Git?").
1673 */
1674 if (insertions || deletions == 0) {
1675 strbuf_addf(&sb,
1676 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
1677 insertions);
1678 }
1679
1680 if (deletions || insertions == 0) {
1681 strbuf_addf(&sb,
1682 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
1683 deletions);
1684 }
1685 strbuf_addch(&sb, '\n');
1686 ret = fputs(sb.buf, fp);
1687 strbuf_release(&sb);
1688 return ret;
1689}
1690
1691static void show_stats(struct diffstat_t *data, struct diff_options *options)
1692{
1693 int i, len, add, del, adds = 0, dels = 0;
1694 uintmax_t max_change = 0, max_len = 0;
1695 int total_files = data->nr, count;
1696 int width, name_width, graph_width, number_width = 0, bin_width = 0;
1697 const char *reset, *add_c, *del_c;
1698 const char *line_prefix = "";
1699 int extra_shown = 0;
1700
1701 if (data->nr == 0)
1702 return;
1703
1704 line_prefix = diff_line_prefix(options);
1705 count = options->stat_count ? options->stat_count : data->nr;
1706
1707 reset = diff_get_color_opt(options, DIFF_RESET);
1708 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1709 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1710
1711 /*
1712 * Find the longest filename and max number of changes
1713 */
1714 for (i = 0; (i < count) && (i < data->nr); i++) {
1715 struct diffstat_file *file = data->files[i];
1716 uintmax_t change = file->added + file->deleted;
1717
1718 if (!file->is_interesting && (change == 0)) {
1719 count++; /* not shown == room for one more */
1720 continue;
1721 }
1722 fill_print_name(file);
1723 len = strlen(file->print_name);
1724 if (max_len < len)
1725 max_len = len;
1726
1727 if (file->is_unmerged) {
1728 /* "Unmerged" is 8 characters */
1729 bin_width = bin_width < 8 ? 8 : bin_width;
1730 continue;
1731 }
1732 if (file->is_binary) {
1733 /* "Bin XXX -> YYY bytes" */
1734 int w = 14 + decimal_width(file->added)
1735 + decimal_width(file->deleted);
1736 bin_width = bin_width < w ? w : bin_width;
1737 /* Display change counts aligned with "Bin" */
1738 number_width = 3;
1739 continue;
1740 }
1741
1742 if (max_change < change)
1743 max_change = change;
1744 }
1745 count = i; /* where we can stop scanning in data->files[] */
1746
1747 /*
1748 * We have width = stat_width or term_columns() columns total.
1749 * We want a maximum of min(max_len, stat_name_width) for the name part.
1750 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
1751 * We also need 1 for " " and 4 + decimal_width(max_change)
1752 * for " | NNNN " and one the empty column at the end, altogether
1753 * 6 + decimal_width(max_change).
1754 *
1755 * If there's not enough space, we will use the smaller of
1756 * stat_name_width (if set) and 5/8*width for the filename,
1757 * and the rest for constant elements + graph part, but no more
1758 * than stat_graph_width for the graph part.
1759 * (5/8 gives 50 for filename and 30 for the constant parts + graph
1760 * for the standard terminal size).
1761 *
1762 * In other words: stat_width limits the maximum width, and
1763 * stat_name_width fixes the maximum width of the filename,
1764 * and is also used to divide available columns if there
1765 * aren't enough.
1766 *
1767 * Binary files are displayed with "Bin XXX -> YYY bytes"
1768 * instead of the change count and graph. This part is treated
1769 * similarly to the graph part, except that it is not
1770 * "scaled". If total width is too small to accommodate the
1771 * guaranteed minimum width of the filename part and the
1772 * separators and this message, this message will "overflow"
1773 * making the line longer than the maximum width.
1774 */
1775
1776 if (options->stat_width == -1)
1777 width = term_columns() - strlen(line_prefix);
1778 else
1779 width = options->stat_width ? options->stat_width : 80;
1780 number_width = decimal_width(max_change) > number_width ?
1781 decimal_width(max_change) : number_width;
1782
1783 if (options->stat_graph_width == -1)
1784 options->stat_graph_width = diff_stat_graph_width;
1785
1786 /*
1787 * Guarantee 3/8*16==6 for the graph part
1788 * and 5/8*16==10 for the filename part
1789 */
1790 if (width < 16 + 6 + number_width)
1791 width = 16 + 6 + number_width;
1792
1793 /*
1794 * First assign sizes that are wanted, ignoring available width.
1795 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
1796 * starting from "XXX" should fit in graph_width.
1797 */
1798 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
1799 if (options->stat_graph_width &&
1800 options->stat_graph_width < graph_width)
1801 graph_width = options->stat_graph_width;
1802
1803 name_width = (options->stat_name_width > 0 &&
1804 options->stat_name_width < max_len) ?
1805 options->stat_name_width : max_len;
1806
1807 /*
1808 * Adjust adjustable widths not to exceed maximum width
1809 */
1810 if (name_width + number_width + 6 + graph_width > width) {
1811 if (graph_width > width * 3/8 - number_width - 6) {
1812 graph_width = width * 3/8 - number_width - 6;
1813 if (graph_width < 6)
1814 graph_width = 6;
1815 }
1816
1817 if (options->stat_graph_width &&
1818 graph_width > options->stat_graph_width)
1819 graph_width = options->stat_graph_width;
1820 if (name_width > width - number_width - 6 - graph_width)
1821 name_width = width - number_width - 6 - graph_width;
1822 else
1823 graph_width = width - number_width - 6 - name_width;
1824 }
1825
1826 /*
1827 * From here name_width is the width of the name area,
1828 * and graph_width is the width of the graph area.
1829 * max_change is used to scale graph properly.
1830 */
1831 for (i = 0; i < count; i++) {
1832 const char *prefix = "";
1833 struct diffstat_file *file = data->files[i];
1834 char *name = file->print_name;
1835 uintmax_t added = file->added;
1836 uintmax_t deleted = file->deleted;
1837 int name_len;
1838
1839 if (!file->is_interesting && (added + deleted == 0))
1840 continue;
1841
1842 /*
1843 * "scale" the filename
1844 */
1845 len = name_width;
1846 name_len = strlen(name);
1847 if (name_width < name_len) {
1848 char *slash;
1849 prefix = "...";
1850 len -= 3;
1851 name += name_len - len;
1852 slash = strchr(name, '/');
1853 if (slash)
1854 name = slash;
1855 }
1856
1857 if (file->is_binary) {
1858 fprintf(options->file, "%s", line_prefix);
1859 show_name(options->file, prefix, name, len);
1860 fprintf(options->file, " %*s", number_width, "Bin");
1861 if (!added && !deleted) {
1862 putc('\n', options->file);
1863 continue;
1864 }
1865 fprintf(options->file, " %s%"PRIuMAX"%s",
1866 del_c, deleted, reset);
1867 fprintf(options->file, " -> ");
1868 fprintf(options->file, "%s%"PRIuMAX"%s",
1869 add_c, added, reset);
1870 fprintf(options->file, " bytes");
1871 fprintf(options->file, "\n");
1872 continue;
1873 }
1874 else if (file->is_unmerged) {
1875 fprintf(options->file, "%s", line_prefix);
1876 show_name(options->file, prefix, name, len);
1877 fprintf(options->file, " Unmerged\n");
1878 continue;
1879 }
1880
1881 /*
1882 * scale the add/delete
1883 */
1884 add = added;
1885 del = deleted;
1886
1887 if (graph_width <= max_change) {
1888 int total = scale_linear(add + del, graph_width, max_change);
1889 if (total < 2 && add && del)
1890 /* width >= 2 due to the sanity check */
1891 total = 2;
1892 if (add < del) {
1893 add = scale_linear(add, graph_width, max_change);
1894 del = total - add;
1895 } else {
1896 del = scale_linear(del, graph_width, max_change);
1897 add = total - del;
1898 }
1899 }
1900 fprintf(options->file, "%s", line_prefix);
1901 show_name(options->file, prefix, name, len);
1902 fprintf(options->file, " %*"PRIuMAX"%s",
1903 number_width, added + deleted,
1904 added + deleted ? " " : "");
1905 show_graph(options->file, '+', add, add_c, reset);
1906 show_graph(options->file, '-', del, del_c, reset);
1907 fprintf(options->file, "\n");
1908 }
1909
1910 for (i = 0; i < data->nr; i++) {
1911 struct diffstat_file *file = data->files[i];
1912 uintmax_t added = file->added;
1913 uintmax_t deleted = file->deleted;
1914
1915 if (file->is_unmerged ||
1916 (!file->is_interesting && (added + deleted == 0))) {
1917 total_files--;
1918 continue;
1919 }
1920
1921 if (!file->is_binary) {
1922 adds += added;
1923 dels += deleted;
1924 }
1925 if (i < count)
1926 continue;
1927 if (!extra_shown)
1928 fprintf(options->file, "%s ...\n", line_prefix);
1929 extra_shown = 1;
1930 }
1931 fprintf(options->file, "%s", line_prefix);
1932 print_stat_summary(options->file, total_files, adds, dels);
1933}
1934
1935static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1936{
1937 int i, adds = 0, dels = 0, total_files = data->nr;
1938
1939 if (data->nr == 0)
1940 return;
1941
1942 for (i = 0; i < data->nr; i++) {
1943 int added = data->files[i]->added;
1944 int deleted= data->files[i]->deleted;
1945
1946 if (data->files[i]->is_unmerged ||
1947 (!data->files[i]->is_interesting && (added + deleted == 0))) {
1948 total_files--;
1949 } else if (!data->files[i]->is_binary) { /* don't count bytes */
1950 adds += added;
1951 dels += deleted;
1952 }
1953 }
1954 fprintf(options->file, "%s", diff_line_prefix(options));
1955 print_stat_summary(options->file, total_files, adds, dels);
1956}
1957
1958static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1959{
1960 int i;
1961
1962 if (data->nr == 0)
1963 return;
1964
1965 for (i = 0; i < data->nr; i++) {
1966 struct diffstat_file *file = data->files[i];
1967
1968 fprintf(options->file, "%s", diff_line_prefix(options));
1969
1970 if (file->is_binary)
1971 fprintf(options->file, "-\t-\t");
1972 else
1973 fprintf(options->file,
1974 "%"PRIuMAX"\t%"PRIuMAX"\t",
1975 file->added, file->deleted);
1976 if (options->line_termination) {
1977 fill_print_name(file);
1978 if (!file->is_renamed)
1979 write_name_quoted(file->name, options->file,
1980 options->line_termination);
1981 else {
1982 fputs(file->print_name, options->file);
1983 putc(options->line_termination, options->file);
1984 }
1985 } else {
1986 if (file->is_renamed) {
1987 putc('\0', options->file);
1988 write_name_quoted(file->from_name, options->file, '\0');
1989 }
1990 write_name_quoted(file->name, options->file, '\0');
1991 }
1992 }
1993}
1994
1995struct dirstat_file {
1996 const char *name;
1997 unsigned long changed;
1998};
1999
2000struct dirstat_dir {
2001 struct dirstat_file *files;
2002 int alloc, nr, permille, cumulative;
2003};
2004
2005static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
2006 unsigned long changed, const char *base, int baselen)
2007{
2008 unsigned long this_dir = 0;
2009 unsigned int sources = 0;
2010 const char *line_prefix = diff_line_prefix(opt);
2011
2012 while (dir->nr) {
2013 struct dirstat_file *f = dir->files;
2014 int namelen = strlen(f->name);
2015 unsigned long this;
2016 char *slash;
2017
2018 if (namelen < baselen)
2019 break;
2020 if (memcmp(f->name, base, baselen))
2021 break;
2022 slash = strchr(f->name + baselen, '/');
2023 if (slash) {
2024 int newbaselen = slash + 1 - f->name;
2025 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
2026 sources++;
2027 } else {
2028 this = f->changed;
2029 dir->files++;
2030 dir->nr--;
2031 sources += 2;
2032 }
2033 this_dir += this;
2034 }
2035
2036 /*
2037 * We don't report dirstat's for
2038 * - the top level
2039 * - or cases where everything came from a single directory
2040 * under this directory (sources == 1).
2041 */
2042 if (baselen && sources != 1) {
2043 if (this_dir) {
2044 int permille = this_dir * 1000 / changed;
2045 if (permille >= dir->permille) {
2046 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
2047 permille / 10, permille % 10, baselen, base);
2048 if (!dir->cumulative)
2049 return 0;
2050 }
2051 }
2052 }
2053 return this_dir;
2054}
2055
2056static int dirstat_compare(const void *_a, const void *_b)
2057{
2058 const struct dirstat_file *a = _a;
2059 const struct dirstat_file *b = _b;
2060 return strcmp(a->name, b->name);
2061}
2062
2063static void show_dirstat(struct diff_options *options)
2064{
2065 int i;
2066 unsigned long changed;
2067 struct dirstat_dir dir;
2068 struct diff_queue_struct *q = &diff_queued_diff;
2069
2070 dir.files = NULL;
2071 dir.alloc = 0;
2072 dir.nr = 0;
2073 dir.permille = options->dirstat_permille;
2074 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2075
2076 changed = 0;
2077 for (i = 0; i < q->nr; i++) {
2078 struct diff_filepair *p = q->queue[i];
2079 const char *name;
2080 unsigned long copied, added, damage;
2081 int content_changed;
2082
2083 name = p->two->path ? p->two->path : p->one->path;
2084
2085 if (p->one->oid_valid && p->two->oid_valid)
2086 content_changed = oidcmp(&p->one->oid, &p->two->oid);
2087 else
2088 content_changed = 1;
2089
2090 if (!content_changed) {
2091 /*
2092 * The SHA1 has not changed, so pre-/post-content is
2093 * identical. We can therefore skip looking at the
2094 * file contents altogether.
2095 */
2096 damage = 0;
2097 goto found_damage;
2098 }
2099
2100 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
2101 /*
2102 * In --dirstat-by-file mode, we don't really need to
2103 * look at the actual file contents at all.
2104 * The fact that the SHA1 changed is enough for us to
2105 * add this file to the list of results
2106 * (with each file contributing equal damage).
2107 */
2108 damage = 1;
2109 goto found_damage;
2110 }
2111
2112 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
2113 diff_populate_filespec(p->one, 0);
2114 diff_populate_filespec(p->two, 0);
2115 diffcore_count_changes(p->one, p->two, NULL, NULL,
2116 &copied, &added);
2117 diff_free_filespec_data(p->one);
2118 diff_free_filespec_data(p->two);
2119 } else if (DIFF_FILE_VALID(p->one)) {
2120 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
2121 copied = added = 0;
2122 diff_free_filespec_data(p->one);
2123 } else if (DIFF_FILE_VALID(p->two)) {
2124 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
2125 copied = 0;
2126 added = p->two->size;
2127 diff_free_filespec_data(p->two);
2128 } else
2129 continue;
2130
2131 /*
2132 * Original minus copied is the removed material,
2133 * added is the new material. They are both damages
2134 * made to the preimage.
2135 * If the resulting damage is zero, we know that
2136 * diffcore_count_changes() considers the two entries to
2137 * be identical, but since content_changed is true, we
2138 * know that there must have been _some_ kind of change,
2139 * so we force all entries to have damage > 0.
2140 */
2141 damage = (p->one->size - copied) + added;
2142 if (!damage)
2143 damage = 1;
2144
2145found_damage:
2146 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2147 dir.files[dir.nr].name = name;
2148 dir.files[dir.nr].changed = damage;
2149 changed += damage;
2150 dir.nr++;
2151 }
2152
2153 /* This can happen even with many files, if everything was renames */
2154 if (!changed)
2155 return;
2156
2157 /* Show all directories with more than x% of the changes */
2158 QSORT(dir.files, dir.nr, dirstat_compare);
2159 gather_dirstat(options, &dir, changed, "", 0);
2160}
2161
2162static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
2163{
2164 int i;
2165 unsigned long changed;
2166 struct dirstat_dir dir;
2167
2168 if (data->nr == 0)
2169 return;
2170
2171 dir.files = NULL;
2172 dir.alloc = 0;
2173 dir.nr = 0;
2174 dir.permille = options->dirstat_permille;
2175 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2176
2177 changed = 0;
2178 for (i = 0; i < data->nr; i++) {
2179 struct diffstat_file *file = data->files[i];
2180 unsigned long damage = file->added + file->deleted;
2181 if (file->is_binary)
2182 /*
2183 * binary files counts bytes, not lines. Must find some
2184 * way to normalize binary bytes vs. textual lines.
2185 * The following heuristic assumes that there are 64
2186 * bytes per "line".
2187 * This is stupid and ugly, but very cheap...
2188 */
2189 damage = (damage + 63) / 64;
2190 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2191 dir.files[dir.nr].name = file->name;
2192 dir.files[dir.nr].changed = damage;
2193 changed += damage;
2194 dir.nr++;
2195 }
2196
2197 /* This can happen even with many files, if everything was renames */
2198 if (!changed)
2199 return;
2200
2201 /* Show all directories with more than x% of the changes */
2202 QSORT(dir.files, dir.nr, dirstat_compare);
2203 gather_dirstat(options, &dir, changed, "", 0);
2204}
2205
2206static void free_diffstat_info(struct diffstat_t *diffstat)
2207{
2208 int i;
2209 for (i = 0; i < diffstat->nr; i++) {
2210 struct diffstat_file *f = diffstat->files[i];
2211 if (f->name != f->print_name)
2212 free(f->print_name);
2213 free(f->name);
2214 free(f->from_name);
2215 free(f);
2216 }
2217 free(diffstat->files);
2218}
2219
2220struct checkdiff_t {
2221 const char *filename;
2222 int lineno;
2223 int conflict_marker_size;
2224 struct diff_options *o;
2225 unsigned ws_rule;
2226 unsigned status;
2227};
2228
2229static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2230{
2231 char firstchar;
2232 int cnt;
2233
2234 if (len < marker_size + 1)
2235 return 0;
2236 firstchar = line[0];
2237 switch (firstchar) {
2238 case '=': case '>': case '<': case '|':
2239 break;
2240 default:
2241 return 0;
2242 }
2243 for (cnt = 1; cnt < marker_size; cnt++)
2244 if (line[cnt] != firstchar)
2245 return 0;
2246 /* line[1] thru line[marker_size-1] are same as firstchar */
2247 if (len < marker_size + 1 || !isspace(line[marker_size]))
2248 return 0;
2249 return 1;
2250}
2251
2252static void checkdiff_consume(void *priv, char *line, unsigned long len)
2253{
2254 struct checkdiff_t *data = priv;
2255 int marker_size = data->conflict_marker_size;
2256 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2257 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2258 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2259 char *err;
2260 const char *line_prefix;
2261
2262 assert(data->o);
2263 line_prefix = diff_line_prefix(data->o);
2264
2265 if (line[0] == '+') {
2266 unsigned bad;
2267 data->lineno++;
2268 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2269 data->status |= 1;
2270 fprintf(data->o->file,
2271 "%s%s:%d: leftover conflict marker\n",
2272 line_prefix, data->filename, data->lineno);
2273 }
2274 bad = ws_check(line + 1, len - 1, data->ws_rule);
2275 if (!bad)
2276 return;
2277 data->status |= bad;
2278 err = whitespace_error_string(bad);
2279 fprintf(data->o->file, "%s%s:%d: %s.\n",
2280 line_prefix, data->filename, data->lineno, err);
2281 free(err);
2282 emit_line(data->o, set, reset, line, 1);
2283 ws_check_emit(line + 1, len - 1, data->ws_rule,
2284 data->o->file, set, reset, ws);
2285 } else if (line[0] == ' ') {
2286 data->lineno++;
2287 } else if (line[0] == '@') {
2288 char *plus = strchr(line, '+');
2289 if (plus)
2290 data->lineno = strtol(plus, NULL, 10) - 1;
2291 else
2292 die("invalid diff");
2293 }
2294}
2295
2296static unsigned char *deflate_it(char *data,
2297 unsigned long size,
2298 unsigned long *result_size)
2299{
2300 int bound;
2301 unsigned char *deflated;
2302 git_zstream stream;
2303
2304 git_deflate_init(&stream, zlib_compression_level);
2305 bound = git_deflate_bound(&stream, size);
2306 deflated = xmalloc(bound);
2307 stream.next_out = deflated;
2308 stream.avail_out = bound;
2309
2310 stream.next_in = (unsigned char *)data;
2311 stream.avail_in = size;
2312 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2313 ; /* nothing */
2314 git_deflate_end(&stream);
2315 *result_size = stream.total_out;
2316 return deflated;
2317}
2318
2319static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two,
2320 const char *prefix)
2321{
2322 void *cp;
2323 void *delta;
2324 void *deflated;
2325 void *data;
2326 unsigned long orig_size;
2327 unsigned long delta_size;
2328 unsigned long deflate_size;
2329 unsigned long data_size;
2330
2331 /* We could do deflated delta, or we could do just deflated two,
2332 * whichever is smaller.
2333 */
2334 delta = NULL;
2335 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2336 if (one->size && two->size) {
2337 delta = diff_delta(one->ptr, one->size,
2338 two->ptr, two->size,
2339 &delta_size, deflate_size);
2340 if (delta) {
2341 void *to_free = delta;
2342 orig_size = delta_size;
2343 delta = deflate_it(delta, delta_size, &delta_size);
2344 free(to_free);
2345 }
2346 }
2347
2348 if (delta && delta_size < deflate_size) {
2349 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
2350 free(deflated);
2351 data = delta;
2352 data_size = delta_size;
2353 }
2354 else {
2355 fprintf(file, "%sliteral %lu\n", prefix, two->size);
2356 free(delta);
2357 data = deflated;
2358 data_size = deflate_size;
2359 }
2360
2361 /* emit data encoded in base85 */
2362 cp = data;
2363 while (data_size) {
2364 int bytes = (52 < data_size) ? 52 : data_size;
2365 char line[70];
2366 data_size -= bytes;
2367 if (bytes <= 26)
2368 line[0] = bytes + 'A' - 1;
2369 else
2370 line[0] = bytes - 26 + 'a' - 1;
2371 encode_85(line + 1, cp, bytes);
2372 cp = (char *) cp + bytes;
2373 fprintf(file, "%s", prefix);
2374 fputs(line, file);
2375 fputc('\n', file);
2376 }
2377 fprintf(file, "%s\n", prefix);
2378 free(data);
2379}
2380
2381static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two,
2382 const char *prefix)
2383{
2384 fprintf(file, "%sGIT binary patch\n", prefix);
2385 emit_binary_diff_body(file, one, two, prefix);
2386 emit_binary_diff_body(file, two, one, prefix);
2387}
2388
2389int diff_filespec_is_binary(struct diff_filespec *one)
2390{
2391 if (one->is_binary == -1) {
2392 diff_filespec_load_driver(one);
2393 if (one->driver->binary != -1)
2394 one->is_binary = one->driver->binary;
2395 else {
2396 if (!one->data && DIFF_FILE_VALID(one))
2397 diff_populate_filespec(one, CHECK_BINARY);
2398 if (one->is_binary == -1 && one->data)
2399 one->is_binary = buffer_is_binary(one->data,
2400 one->size);
2401 if (one->is_binary == -1)
2402 one->is_binary = 0;
2403 }
2404 }
2405 return one->is_binary;
2406}
2407
2408static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2409{
2410 diff_filespec_load_driver(one);
2411 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2412}
2413
2414void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2415{
2416 if (!options->a_prefix)
2417 options->a_prefix = a;
2418 if (!options->b_prefix)
2419 options->b_prefix = b;
2420}
2421
2422struct userdiff_driver *get_textconv(struct diff_filespec *one)
2423{
2424 if (!DIFF_FILE_VALID(one))
2425 return NULL;
2426
2427 diff_filespec_load_driver(one);
2428 return userdiff_get_textconv(one->driver);
2429}
2430
2431static void builtin_diff(const char *name_a,
2432 const char *name_b,
2433 struct diff_filespec *one,
2434 struct diff_filespec *two,
2435 const char *xfrm_msg,
2436 int must_show_header,
2437 struct diff_options *o,
2438 int complete_rewrite)
2439{
2440 mmfile_t mf1, mf2;
2441 const char *lbl[2];
2442 char *a_one, *b_two;
2443 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
2444 const char *reset = diff_get_color_opt(o, DIFF_RESET);
2445 const char *a_prefix, *b_prefix;
2446 struct userdiff_driver *textconv_one = NULL;
2447 struct userdiff_driver *textconv_two = NULL;
2448 struct strbuf header = STRBUF_INIT;
2449 const char *line_prefix = diff_line_prefix(o);
2450
2451 diff_set_mnemonic_prefix(o, "a/", "b/");
2452 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2453 a_prefix = o->b_prefix;
2454 b_prefix = o->a_prefix;
2455 } else {
2456 a_prefix = o->a_prefix;
2457 b_prefix = o->b_prefix;
2458 }
2459
2460 if (o->submodule_format == DIFF_SUBMODULE_LOG &&
2461 (!one->mode || S_ISGITLINK(one->mode)) &&
2462 (!two->mode || S_ISGITLINK(two->mode))) {
2463 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2464 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2465 show_submodule_summary(o->file, one->path ? one->path : two->path,
2466 line_prefix,
2467 &one->oid, &two->oid,
2468 two->dirty_submodule,
2469 meta, del, add, reset);
2470 return;
2471 } else if (o->submodule_format == DIFF_SUBMODULE_INLINE_DIFF &&
2472 (!one->mode || S_ISGITLINK(one->mode)) &&
2473 (!two->mode || S_ISGITLINK(two->mode))) {
2474 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2475 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2476 show_submodule_inline_diff(o->file, one->path ? one->path : two->path,
2477 line_prefix,
2478 &one->oid, &two->oid,
2479 two->dirty_submodule,
2480 meta, del, add, reset, o);
2481 return;
2482 }
2483
2484 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2485 textconv_one = get_textconv(one);
2486 textconv_two = get_textconv(two);
2487 }
2488
2489 /* Never use a non-valid filename anywhere if at all possible */
2490 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2491 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2492
2493 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2494 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2495 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2496 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2497 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
2498 if (lbl[0][0] == '/') {
2499 /* /dev/null */
2500 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
2501 if (xfrm_msg)
2502 strbuf_addstr(&header, xfrm_msg);
2503 must_show_header = 1;
2504 }
2505 else if (lbl[1][0] == '/') {
2506 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
2507 if (xfrm_msg)
2508 strbuf_addstr(&header, xfrm_msg);
2509 must_show_header = 1;
2510 }
2511 else {
2512 if (one->mode != two->mode) {
2513 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
2514 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
2515 must_show_header = 1;
2516 }
2517 if (xfrm_msg)
2518 strbuf_addstr(&header, xfrm_msg);
2519
2520 /*
2521 * we do not run diff between different kind
2522 * of objects.
2523 */
2524 if ((one->mode ^ two->mode) & S_IFMT)
2525 goto free_ab_and_return;
2526 if (complete_rewrite &&
2527 (textconv_one || !diff_filespec_is_binary(one)) &&
2528 (textconv_two || !diff_filespec_is_binary(two))) {
2529 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
2530 header.buf, header.len, 0);
2531 strbuf_reset(&header);
2532 emit_rewrite_diff(name_a, name_b, one, two,
2533 textconv_one, textconv_two, o);
2534 o->found_changes = 1;
2535 goto free_ab_and_return;
2536 }
2537 }
2538
2539 if (o->irreversible_delete && lbl[1][0] == '/') {
2540 emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf,
2541 header.len, 0);
2542 strbuf_reset(&header);
2543 goto free_ab_and_return;
2544 } else if (!DIFF_OPT_TST(o, TEXT) &&
2545 ( (!textconv_one && diff_filespec_is_binary(one)) ||
2546 (!textconv_two && diff_filespec_is_binary(two)) )) {
2547 struct strbuf sb = STRBUF_INIT;
2548 if (!one->data && !two->data &&
2549 S_ISREG(one->mode) && S_ISREG(two->mode) &&
2550 !DIFF_OPT_TST(o, BINARY)) {
2551 if (!oidcmp(&one->oid, &two->oid)) {
2552 if (must_show_header)
2553 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
2554 header.buf, header.len,
2555 0);
2556 goto free_ab_and_return;
2557 }
2558 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
2559 header.buf, header.len, 0);
2560 strbuf_addf(&sb, "%sBinary files %s and %s differ\n",
2561 diff_line_prefix(o), lbl[0], lbl[1]);
2562 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_FILES,
2563 sb.buf, sb.len, 0);
2564 strbuf_release(&sb);
2565 goto free_ab_and_return;
2566 }
2567 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2568 die("unable to read files to diff");
2569 /* Quite common confusing case */
2570 if (mf1.size == mf2.size &&
2571 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2572 if (must_show_header)
2573 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
2574 header.buf, header.len, 0);
2575 goto free_ab_and_return;
2576 }
2577 emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf, header.len, 0);
2578 strbuf_reset(&header);
2579 if (DIFF_OPT_TST(o, BINARY))
2580 emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2581 else {
2582 strbuf_addf(&sb, "%sBinary files %s and %s differ\n",
2583 diff_line_prefix(o), lbl[0], lbl[1]);
2584 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_FILES,
2585 sb.buf, sb.len, 0);
2586 strbuf_release(&sb);
2587 }
2588 o->found_changes = 1;
2589 } else {
2590 /* Crazy xdl interfaces.. */
2591 const char *diffopts = getenv("GIT_DIFF_OPTS");
2592 const char *v;
2593 xpparam_t xpp;
2594 xdemitconf_t xecfg;
2595 struct emit_callback ecbdata;
2596 const struct userdiff_funcname *pe;
2597
2598 if (must_show_header) {
2599 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
2600 header.buf, header.len, 0);
2601 strbuf_reset(&header);
2602 }
2603
2604 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2605 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2606
2607 pe = diff_funcname_pattern(one);
2608 if (!pe)
2609 pe = diff_funcname_pattern(two);
2610
2611 memset(&xpp, 0, sizeof(xpp));
2612 memset(&xecfg, 0, sizeof(xecfg));
2613 memset(&ecbdata, 0, sizeof(ecbdata));
2614 ecbdata.label_path = lbl;
2615 ecbdata.color_diff = want_color(o->use_color);
2616 ecbdata.ws_rule = whitespace_rule(name_b);
2617 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2618 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2619 ecbdata.opt = o;
2620 ecbdata.header = header.len ? &header : NULL;
2621 xpp.flags = o->xdl_opts;
2622 xecfg.ctxlen = o->context;
2623 xecfg.interhunkctxlen = o->interhunkcontext;
2624 xecfg.flags = XDL_EMIT_FUNCNAMES;
2625 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2626 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2627 if (pe)
2628 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2629 if (!diffopts)
2630 ;
2631 else if (skip_prefix(diffopts, "--unified=", &v))
2632 xecfg.ctxlen = strtoul(v, NULL, 10);
2633 else if (skip_prefix(diffopts, "-u", &v))
2634 xecfg.ctxlen = strtoul(v, NULL, 10);
2635 if (o->word_diff)
2636 init_diff_words_data(&ecbdata, o, one, two);
2637 if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2638 &xpp, &xecfg))
2639 die("unable to generate diff for %s", one->path);
2640 if (o->word_diff)
2641 free_diff_words_data(&ecbdata);
2642 if (textconv_one)
2643 free(mf1.ptr);
2644 if (textconv_two)
2645 free(mf2.ptr);
2646 xdiff_clear_find_func(&xecfg);
2647 }
2648
2649 free_ab_and_return:
2650 strbuf_release(&header);
2651 diff_free_filespec_data(one);
2652 diff_free_filespec_data(two);
2653 free(a_one);
2654 free(b_two);
2655 return;
2656}
2657
2658static void builtin_diffstat(const char *name_a, const char *name_b,
2659 struct diff_filespec *one,
2660 struct diff_filespec *two,
2661 struct diffstat_t *diffstat,
2662 struct diff_options *o,
2663 struct diff_filepair *p)
2664{
2665 mmfile_t mf1, mf2;
2666 struct diffstat_file *data;
2667 int same_contents;
2668 int complete_rewrite = 0;
2669
2670 if (!DIFF_PAIR_UNMERGED(p)) {
2671 if (p->status == DIFF_STATUS_MODIFIED && p->score)
2672 complete_rewrite = 1;
2673 }
2674
2675 data = diffstat_add(diffstat, name_a, name_b);
2676 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
2677
2678 if (!one || !two) {
2679 data->is_unmerged = 1;
2680 return;
2681 }
2682
2683 same_contents = !oidcmp(&one->oid, &two->oid);
2684
2685 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2686 data->is_binary = 1;
2687 if (same_contents) {
2688 data->added = 0;
2689 data->deleted = 0;
2690 } else {
2691 data->added = diff_filespec_size(two);
2692 data->deleted = diff_filespec_size(one);
2693 }
2694 }
2695
2696 else if (complete_rewrite) {
2697 diff_populate_filespec(one, 0);
2698 diff_populate_filespec(two, 0);
2699 data->deleted = count_lines(one->data, one->size);
2700 data->added = count_lines(two->data, two->size);
2701 }
2702
2703 else if (!same_contents) {
2704 /* Crazy xdl interfaces.. */
2705 xpparam_t xpp;
2706 xdemitconf_t xecfg;
2707
2708 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2709 die("unable to read files to diff");
2710
2711 memset(&xpp, 0, sizeof(xpp));
2712 memset(&xecfg, 0, sizeof(xecfg));
2713 xpp.flags = o->xdl_opts;
2714 xecfg.ctxlen = o->context;
2715 xecfg.interhunkctxlen = o->interhunkcontext;
2716 if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2717 &xpp, &xecfg))
2718 die("unable to generate diffstat for %s", one->path);
2719 }
2720
2721 diff_free_filespec_data(one);
2722 diff_free_filespec_data(two);
2723}
2724
2725static void builtin_checkdiff(const char *name_a, const char *name_b,
2726 const char *attr_path,
2727 struct diff_filespec *one,
2728 struct diff_filespec *two,
2729 struct diff_options *o)
2730{
2731 mmfile_t mf1, mf2;
2732 struct checkdiff_t data;
2733
2734 if (!two)
2735 return;
2736
2737 memset(&data, 0, sizeof(data));
2738 data.filename = name_b ? name_b : name_a;
2739 data.lineno = 0;
2740 data.o = o;
2741 data.ws_rule = whitespace_rule(attr_path);
2742 data.conflict_marker_size = ll_merge_marker_size(attr_path);
2743
2744 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2745 die("unable to read files to diff");
2746
2747 /*
2748 * All the other codepaths check both sides, but not checking
2749 * the "old" side here is deliberate. We are checking the newly
2750 * introduced changes, and as long as the "new" side is text, we
2751 * can and should check what it introduces.
2752 */
2753 if (diff_filespec_is_binary(two))
2754 goto free_and_return;
2755 else {
2756 /* Crazy xdl interfaces.. */
2757 xpparam_t xpp;
2758 xdemitconf_t xecfg;
2759
2760 memset(&xpp, 0, sizeof(xpp));
2761 memset(&xecfg, 0, sizeof(xecfg));
2762 xecfg.ctxlen = 1; /* at least one context line */
2763 xpp.flags = 0;
2764 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2765 &xpp, &xecfg))
2766 die("unable to generate checkdiff for %s", one->path);
2767
2768 if (data.ws_rule & WS_BLANK_AT_EOF) {
2769 struct emit_callback ecbdata;
2770 int blank_at_eof;
2771
2772 ecbdata.ws_rule = data.ws_rule;
2773 check_blank_at_eof(&mf1, &mf2, &ecbdata);
2774 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2775
2776 if (blank_at_eof) {
2777 static char *err;
2778 if (!err)
2779 err = whitespace_error_string(WS_BLANK_AT_EOF);
2780 fprintf(o->file, "%s:%d: %s.\n",
2781 data.filename, blank_at_eof, err);
2782 data.status = 1; /* report errors */
2783 }
2784 }
2785 }
2786 free_and_return:
2787 diff_free_filespec_data(one);
2788 diff_free_filespec_data(two);
2789 if (data.status)
2790 DIFF_OPT_SET(o, CHECK_FAILED);
2791}
2792
2793struct diff_filespec *alloc_filespec(const char *path)
2794{
2795 struct diff_filespec *spec;
2796
2797 FLEXPTR_ALLOC_STR(spec, path, path);
2798 spec->count = 1;
2799 spec->is_binary = -1;
2800 return spec;
2801}
2802
2803void free_filespec(struct diff_filespec *spec)
2804{
2805 if (!--spec->count) {
2806 diff_free_filespec_data(spec);
2807 free(spec);
2808 }
2809}
2810
2811void fill_filespec(struct diff_filespec *spec, const struct object_id *oid,
2812 int oid_valid, unsigned short mode)
2813{
2814 if (mode) {
2815 spec->mode = canon_mode(mode);
2816 oidcpy(&spec->oid, oid);
2817 spec->oid_valid = oid_valid;
2818 }
2819}
2820
2821/*
2822 * Given a name and sha1 pair, if the index tells us the file in
2823 * the work tree has that object contents, return true, so that
2824 * prepare_temp_file() does not have to inflate and extract.
2825 */
2826static int reuse_worktree_file(const char *name, const struct object_id *oid, int want_file)
2827{
2828 const struct cache_entry *ce;
2829 struct stat st;
2830 int pos, len;
2831
2832 /*
2833 * We do not read the cache ourselves here, because the
2834 * benchmark with my previous version that always reads cache
2835 * shows that it makes things worse for diff-tree comparing
2836 * two linux-2.6 kernel trees in an already checked out work
2837 * tree. This is because most diff-tree comparisons deal with
2838 * only a small number of files, while reading the cache is
2839 * expensive for a large project, and its cost outweighs the
2840 * savings we get by not inflating the object to a temporary
2841 * file. Practically, this code only helps when we are used
2842 * by diff-cache --cached, which does read the cache before
2843 * calling us.
2844 */
2845 if (!active_cache)
2846 return 0;
2847
2848 /* We want to avoid the working directory if our caller
2849 * doesn't need the data in a normal file, this system
2850 * is rather slow with its stat/open/mmap/close syscalls,
2851 * and the object is contained in a pack file. The pack
2852 * is probably already open and will be faster to obtain
2853 * the data through than the working directory. Loose
2854 * objects however would tend to be slower as they need
2855 * to be individually opened and inflated.
2856 */
2857 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(oid->hash))
2858 return 0;
2859
2860 /*
2861 * Similarly, if we'd have to convert the file contents anyway, that
2862 * makes the optimization not worthwhile.
2863 */
2864 if (!want_file && would_convert_to_git(&the_index, name))
2865 return 0;
2866
2867 len = strlen(name);
2868 pos = cache_name_pos(name, len);
2869 if (pos < 0)
2870 return 0;
2871 ce = active_cache[pos];
2872
2873 /*
2874 * This is not the sha1 we are looking for, or
2875 * unreusable because it is not a regular file.
2876 */
2877 if (oidcmp(oid, &ce->oid) || !S_ISREG(ce->ce_mode))
2878 return 0;
2879
2880 /*
2881 * If ce is marked as "assume unchanged", there is no
2882 * guarantee that work tree matches what we are looking for.
2883 */
2884 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2885 return 0;
2886
2887 /*
2888 * If ce matches the file in the work tree, we can reuse it.
2889 */
2890 if (ce_uptodate(ce) ||
2891 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2892 return 1;
2893
2894 return 0;
2895}
2896
2897static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2898{
2899 struct strbuf buf = STRBUF_INIT;
2900 char *dirty = "";
2901
2902 /* Are we looking at the work tree? */
2903 if (s->dirty_submodule)
2904 dirty = "-dirty";
2905
2906 strbuf_addf(&buf, "Subproject commit %s%s\n",
2907 oid_to_hex(&s->oid), dirty);
2908 s->size = buf.len;
2909 if (size_only) {
2910 s->data = NULL;
2911 strbuf_release(&buf);
2912 } else {
2913 s->data = strbuf_detach(&buf, NULL);
2914 s->should_free = 1;
2915 }
2916 return 0;
2917}
2918
2919/*
2920 * While doing rename detection and pickaxe operation, we may need to
2921 * grab the data for the blob (or file) for our own in-core comparison.
2922 * diff_filespec has data and size fields for this purpose.
2923 */
2924int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
2925{
2926 int size_only = flags & CHECK_SIZE_ONLY;
2927 int err = 0;
2928 /*
2929 * demote FAIL to WARN to allow inspecting the situation
2930 * instead of refusing.
2931 */
2932 enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
2933 ? SAFE_CRLF_WARN
2934 : safe_crlf);
2935
2936 if (!DIFF_FILE_VALID(s))
2937 die("internal error: asking to populate invalid file.");
2938 if (S_ISDIR(s->mode))
2939 return -1;
2940
2941 if (s->data)
2942 return 0;
2943
2944 if (size_only && 0 < s->size)
2945 return 0;
2946
2947 if (S_ISGITLINK(s->mode))
2948 return diff_populate_gitlink(s, size_only);
2949
2950 if (!s->oid_valid ||
2951 reuse_worktree_file(s->path, &s->oid, 0)) {
2952 struct strbuf buf = STRBUF_INIT;
2953 struct stat st;
2954 int fd;
2955
2956 if (lstat(s->path, &st) < 0) {
2957 if (errno == ENOENT) {
2958 err_empty:
2959 err = -1;
2960 empty:
2961 s->data = (char *)"";
2962 s->size = 0;
2963 return err;
2964 }
2965 }
2966 s->size = xsize_t(st.st_size);
2967 if (!s->size)
2968 goto empty;
2969 if (S_ISLNK(st.st_mode)) {
2970 struct strbuf sb = STRBUF_INIT;
2971
2972 if (strbuf_readlink(&sb, s->path, s->size))
2973 goto err_empty;
2974 s->size = sb.len;
2975 s->data = strbuf_detach(&sb, NULL);
2976 s->should_free = 1;
2977 return 0;
2978 }
2979
2980 /*
2981 * Even if the caller would be happy with getting
2982 * only the size, we cannot return early at this
2983 * point if the path requires us to run the content
2984 * conversion.
2985 */
2986 if (size_only && !would_convert_to_git(&the_index, s->path))
2987 return 0;
2988
2989 /*
2990 * Note: this check uses xsize_t(st.st_size) that may
2991 * not be the true size of the blob after it goes
2992 * through convert_to_git(). This may not strictly be
2993 * correct, but the whole point of big_file_threshold
2994 * and is_binary check being that we want to avoid
2995 * opening the file and inspecting the contents, this
2996 * is probably fine.
2997 */
2998 if ((flags & CHECK_BINARY) &&
2999 s->size > big_file_threshold && s->is_binary == -1) {
3000 s->is_binary = 1;
3001 return 0;
3002 }
3003 fd = open(s->path, O_RDONLY);
3004 if (fd < 0)
3005 goto err_empty;
3006 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
3007 close(fd);
3008 s->should_munmap = 1;
3009
3010 /*
3011 * Convert from working tree format to canonical git format
3012 */
3013 if (convert_to_git(&the_index, s->path, s->data, s->size, &buf, crlf_warn)) {
3014 size_t size = 0;
3015 munmap(s->data, s->size);
3016 s->should_munmap = 0;
3017 s->data = strbuf_detach(&buf, &size);
3018 s->size = size;
3019 s->should_free = 1;
3020 }
3021 }
3022 else {
3023 enum object_type type;
3024 if (size_only || (flags & CHECK_BINARY)) {
3025 type = sha1_object_info(s->oid.hash, &s->size);
3026 if (type < 0)
3027 die("unable to read %s",
3028 oid_to_hex(&s->oid));
3029 if (size_only)
3030 return 0;
3031 if (s->size > big_file_threshold && s->is_binary == -1) {
3032 s->is_binary = 1;
3033 return 0;
3034 }
3035 }
3036 s->data = read_sha1_file(s->oid.hash, &type, &s->size);
3037 if (!s->data)
3038 die("unable to read %s", oid_to_hex(&s->oid));
3039 s->should_free = 1;
3040 }
3041 return 0;
3042}
3043
3044void diff_free_filespec_blob(struct diff_filespec *s)
3045{
3046 if (s->should_free)
3047 free(s->data);
3048 else if (s->should_munmap)
3049 munmap(s->data, s->size);
3050
3051 if (s->should_free || s->should_munmap) {
3052 s->should_free = s->should_munmap = 0;
3053 s->data = NULL;
3054 }
3055}
3056
3057void diff_free_filespec_data(struct diff_filespec *s)
3058{
3059 diff_free_filespec_blob(s);
3060 FREE_AND_NULL(s->cnt_data);
3061}
3062
3063static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
3064 void *blob,
3065 unsigned long size,
3066 const struct object_id *oid,
3067 int mode)
3068{
3069 int fd;
3070 struct strbuf buf = STRBUF_INIT;
3071 struct strbuf template = STRBUF_INIT;
3072 char *path_dup = xstrdup(path);
3073 const char *base = basename(path_dup);
3074
3075 /* Generate "XXXXXX_basename.ext" */
3076 strbuf_addstr(&template, "XXXXXX_");
3077 strbuf_addstr(&template, base);
3078
3079 fd = mks_tempfile_ts(&temp->tempfile, template.buf, strlen(base) + 1);
3080 if (fd < 0)
3081 die_errno("unable to create temp-file");
3082 if (convert_to_working_tree(path,
3083 (const char *)blob, (size_t)size, &buf)) {
3084 blob = buf.buf;
3085 size = buf.len;
3086 }
3087 if (write_in_full(fd, blob, size) != size)
3088 die_errno("unable to write temp-file");
3089 close_tempfile(&temp->tempfile);
3090 temp->name = get_tempfile_path(&temp->tempfile);
3091 oid_to_hex_r(temp->hex, oid);
3092 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", mode);
3093 strbuf_release(&buf);
3094 strbuf_release(&template);
3095 free(path_dup);
3096}
3097
3098static struct diff_tempfile *prepare_temp_file(const char *name,
3099 struct diff_filespec *one)
3100{
3101 struct diff_tempfile *temp = claim_diff_tempfile();
3102
3103 if (!DIFF_FILE_VALID(one)) {
3104 not_a_valid_file:
3105 /* A '-' entry produces this for file-2, and
3106 * a '+' entry produces this for file-1.
3107 */
3108 temp->name = "/dev/null";
3109 xsnprintf(temp->hex, sizeof(temp->hex), ".");
3110 xsnprintf(temp->mode, sizeof(temp->mode), ".");
3111 return temp;
3112 }
3113
3114 if (!S_ISGITLINK(one->mode) &&
3115 (!one->oid_valid ||
3116 reuse_worktree_file(name, &one->oid, 1))) {
3117 struct stat st;
3118 if (lstat(name, &st) < 0) {
3119 if (errno == ENOENT)
3120 goto not_a_valid_file;
3121 die_errno("stat(%s)", name);
3122 }
3123 if (S_ISLNK(st.st_mode)) {
3124 struct strbuf sb = STRBUF_INIT;
3125 if (strbuf_readlink(&sb, name, st.st_size) < 0)
3126 die_errno("readlink(%s)", name);
3127 prep_temp_blob(name, temp, sb.buf, sb.len,
3128 (one->oid_valid ?
3129 &one->oid : &null_oid),
3130 (one->oid_valid ?
3131 one->mode : S_IFLNK));
3132 strbuf_release(&sb);
3133 }
3134 else {
3135 /* we can borrow from the file in the work tree */
3136 temp->name = name;
3137 if (!one->oid_valid)
3138 oid_to_hex_r(temp->hex, &null_oid);
3139 else
3140 oid_to_hex_r(temp->hex, &one->oid);
3141 /* Even though we may sometimes borrow the
3142 * contents from the work tree, we always want
3143 * one->mode. mode is trustworthy even when
3144 * !(one->oid_valid), as long as
3145 * DIFF_FILE_VALID(one).
3146 */
3147 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", one->mode);
3148 }
3149 return temp;
3150 }
3151 else {
3152 if (diff_populate_filespec(one, 0))
3153 die("cannot read data blob for %s", one->path);
3154 prep_temp_blob(name, temp, one->data, one->size,
3155 &one->oid, one->mode);
3156 }
3157 return temp;
3158}
3159
3160static void add_external_diff_name(struct argv_array *argv,
3161 const char *name,
3162 struct diff_filespec *df)
3163{
3164 struct diff_tempfile *temp = prepare_temp_file(name, df);
3165 argv_array_push(argv, temp->name);
3166 argv_array_push(argv, temp->hex);
3167 argv_array_push(argv, temp->mode);
3168}
3169
3170/* An external diff command takes:
3171 *
3172 * diff-cmd name infile1 infile1-sha1 infile1-mode \
3173 * infile2 infile2-sha1 infile2-mode [ rename-to ]
3174 *
3175 */
3176static void run_external_diff(const char *pgm,
3177 const char *name,
3178 const char *other,
3179 struct diff_filespec *one,
3180 struct diff_filespec *two,
3181 const char *xfrm_msg,
3182 int complete_rewrite,
3183 struct diff_options *o)
3184{
3185 struct argv_array argv = ARGV_ARRAY_INIT;
3186 struct argv_array env = ARGV_ARRAY_INIT;
3187 struct diff_queue_struct *q = &diff_queued_diff;
3188
3189 argv_array_push(&argv, pgm);
3190 argv_array_push(&argv, name);
3191
3192 if (one && two) {
3193 add_external_diff_name(&argv, name, one);
3194 if (!other)
3195 add_external_diff_name(&argv, name, two);
3196 else {
3197 add_external_diff_name(&argv, other, two);
3198 argv_array_push(&argv, other);
3199 argv_array_push(&argv, xfrm_msg);
3200 }
3201 }
3202
3203 argv_array_pushf(&env, "GIT_DIFF_PATH_COUNTER=%d", ++o->diff_path_counter);
3204 argv_array_pushf(&env, "GIT_DIFF_PATH_TOTAL=%d", q->nr);
3205
3206 if (run_command_v_opt_cd_env(argv.argv, RUN_USING_SHELL, NULL, env.argv))
3207 die(_("external diff died, stopping at %s"), name);
3208
3209 remove_tempfile();
3210 argv_array_clear(&argv);
3211 argv_array_clear(&env);
3212}
3213
3214static int similarity_index(struct diff_filepair *p)
3215{
3216 return p->score * 100 / MAX_SCORE;
3217}
3218
3219static const char *diff_abbrev_oid(const struct object_id *oid, int abbrev)
3220{
3221 if (startup_info->have_repository)
3222 return find_unique_abbrev(oid->hash, abbrev);
3223 else {
3224 char *hex = oid_to_hex(oid);
3225 if (abbrev < 0)
3226 abbrev = FALLBACK_DEFAULT_ABBREV;
3227 if (abbrev > GIT_SHA1_HEXSZ)
3228 die("BUG: oid abbreviation out of range: %d", abbrev);
3229 if (abbrev)
3230 hex[abbrev] = '\0';
3231 return hex;
3232 }
3233}
3234
3235static void fill_metainfo(struct strbuf *msg,
3236 const char *name,
3237 const char *other,
3238 struct diff_filespec *one,
3239 struct diff_filespec *two,
3240 struct diff_options *o,
3241 struct diff_filepair *p,
3242 int *must_show_header,
3243 int use_color)
3244{
3245 const char *set = diff_get_color(use_color, DIFF_METAINFO);
3246 const char *reset = diff_get_color(use_color, DIFF_RESET);
3247 const char *line_prefix = diff_line_prefix(o);
3248
3249 *must_show_header = 1;
3250 strbuf_init(msg, PATH_MAX * 2 + 300);
3251 switch (p->status) {
3252 case DIFF_STATUS_COPIED:
3253 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3254 line_prefix, set, similarity_index(p));
3255 strbuf_addf(msg, "%s\n%s%scopy from ",
3256 reset, line_prefix, set);
3257 quote_c_style(name, msg, NULL, 0);
3258 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
3259 quote_c_style(other, msg, NULL, 0);
3260 strbuf_addf(msg, "%s\n", reset);
3261 break;
3262 case DIFF_STATUS_RENAMED:
3263 strbuf_addf(msg, "%s%ssimilarity index %d%%",
3264 line_prefix, set, similarity_index(p));
3265 strbuf_addf(msg, "%s\n%s%srename from ",
3266 reset, line_prefix, set);
3267 quote_c_style(name, msg, NULL, 0);
3268 strbuf_addf(msg, "%s\n%s%srename to ",
3269 reset, line_prefix, set);
3270 quote_c_style(other, msg, NULL, 0);
3271 strbuf_addf(msg, "%s\n", reset);
3272 break;
3273 case DIFF_STATUS_MODIFIED:
3274 if (p->score) {
3275 strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
3276 line_prefix,
3277 set, similarity_index(p), reset);
3278 break;
3279 }
3280 /* fallthru */
3281 default:
3282 *must_show_header = 0;
3283 }
3284 if (one && two && oidcmp(&one->oid, &two->oid)) {
3285 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
3286
3287 if (DIFF_OPT_TST(o, BINARY)) {
3288 mmfile_t mf;
3289 if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
3290 (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
3291 abbrev = 40;
3292 }
3293 strbuf_addf(msg, "%s%sindex %s..%s", line_prefix, set,
3294 diff_abbrev_oid(&one->oid, abbrev),
3295 diff_abbrev_oid(&two->oid, abbrev));
3296 if (one->mode == two->mode)
3297 strbuf_addf(msg, " %06o", one->mode);
3298 strbuf_addf(msg, "%s\n", reset);
3299 }
3300}
3301
3302static void run_diff_cmd(const char *pgm,
3303 const char *name,
3304 const char *other,
3305 const char *attr_path,
3306 struct diff_filespec *one,
3307 struct diff_filespec *two,
3308 struct strbuf *msg,
3309 struct diff_options *o,
3310 struct diff_filepair *p)
3311{
3312 const char *xfrm_msg = NULL;
3313 int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
3314 int must_show_header = 0;
3315
3316
3317 if (DIFF_OPT_TST(o, ALLOW_EXTERNAL)) {
3318 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
3319 if (drv && drv->external)
3320 pgm = drv->external;
3321 }
3322
3323 if (msg) {
3324 /*
3325 * don't use colors when the header is intended for an
3326 * external diff driver
3327 */
3328 fill_metainfo(msg, name, other, one, two, o, p,
3329 &must_show_header,
3330 want_color(o->use_color) && !pgm);
3331 xfrm_msg = msg->len ? msg->buf : NULL;
3332 }
3333
3334 if (pgm) {
3335 run_external_diff(pgm, name, other, one, two, xfrm_msg,
3336 complete_rewrite, o);
3337 return;
3338 }
3339 if (one && two)
3340 builtin_diff(name, other ? other : name,
3341 one, two, xfrm_msg, must_show_header,
3342 o, complete_rewrite);
3343 else
3344 fprintf(o->file, "* Unmerged path %s\n", name);
3345}
3346
3347static void diff_fill_oid_info(struct diff_filespec *one)
3348{
3349 if (DIFF_FILE_VALID(one)) {
3350 if (!one->oid_valid) {
3351 struct stat st;
3352 if (one->is_stdin) {
3353 oidclr(&one->oid);
3354 return;
3355 }
3356 if (lstat(one->path, &st) < 0)
3357 die_errno("stat '%s'", one->path);
3358 if (index_path(one->oid.hash, one->path, &st, 0))
3359 die("cannot hash %s", one->path);
3360 }
3361 }
3362 else
3363 oidclr(&one->oid);
3364}
3365
3366static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
3367{
3368 /* Strip the prefix but do not molest /dev/null and absolute paths */
3369 if (*namep && **namep != '/') {
3370 *namep += prefix_length;
3371 if (**namep == '/')
3372 ++*namep;
3373 }
3374 if (*otherp && **otherp != '/') {
3375 *otherp += prefix_length;
3376 if (**otherp == '/')
3377 ++*otherp;
3378 }
3379}
3380
3381static void run_diff(struct diff_filepair *p, struct diff_options *o)
3382{
3383 const char *pgm = external_diff();
3384 struct strbuf msg;
3385 struct diff_filespec *one = p->one;
3386 struct diff_filespec *two = p->two;
3387 const char *name;
3388 const char *other;
3389 const char *attr_path;
3390
3391 name = one->path;
3392 other = (strcmp(name, two->path) ? two->path : NULL);
3393 attr_path = name;
3394 if (o->prefix_length)
3395 strip_prefix(o->prefix_length, &name, &other);
3396
3397 if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
3398 pgm = NULL;
3399
3400 if (DIFF_PAIR_UNMERGED(p)) {
3401 run_diff_cmd(pgm, name, NULL, attr_path,
3402 NULL, NULL, NULL, o, p);
3403 return;
3404 }
3405
3406 diff_fill_oid_info(one);
3407 diff_fill_oid_info(two);
3408
3409 if (!pgm &&
3410 DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
3411 (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
3412 /*
3413 * a filepair that changes between file and symlink
3414 * needs to be split into deletion and creation.
3415 */
3416 struct diff_filespec *null = alloc_filespec(two->path);
3417 run_diff_cmd(NULL, name, other, attr_path,
3418 one, null, &msg, o, p);
3419 free(null);
3420 strbuf_release(&msg);
3421
3422 null = alloc_filespec(one->path);
3423 run_diff_cmd(NULL, name, other, attr_path,
3424 null, two, &msg, o, p);
3425 free(null);
3426 }
3427 else
3428 run_diff_cmd(pgm, name, other, attr_path,
3429 one, two, &msg, o, p);
3430
3431 strbuf_release(&msg);
3432}
3433
3434static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
3435 struct diffstat_t *diffstat)
3436{
3437 const char *name;
3438 const char *other;
3439
3440 if (DIFF_PAIR_UNMERGED(p)) {
3441 /* unmerged */
3442 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, p);
3443 return;
3444 }
3445
3446 name = p->one->path;
3447 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3448
3449 if (o->prefix_length)
3450 strip_prefix(o->prefix_length, &name, &other);
3451
3452 diff_fill_oid_info(p->one);
3453 diff_fill_oid_info(p->two);
3454
3455 builtin_diffstat(name, other, p->one, p->two, diffstat, o, p);
3456}
3457
3458static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
3459{
3460 const char *name;
3461 const char *other;
3462 const char *attr_path;
3463
3464 if (DIFF_PAIR_UNMERGED(p)) {
3465 /* unmerged */
3466 return;
3467 }
3468
3469 name = p->one->path;
3470 other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3471 attr_path = other ? other : name;
3472
3473 if (o->prefix_length)
3474 strip_prefix(o->prefix_length, &name, &other);
3475
3476 diff_fill_oid_info(p->one);
3477 diff_fill_oid_info(p->two);
3478
3479 builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3480}
3481
3482void diff_setup(struct diff_options *options)
3483{
3484 memcpy(options, &default_diff_options, sizeof(*options));
3485
3486 options->file = stdout;
3487
3488 options->abbrev = DEFAULT_ABBREV;
3489 options->line_termination = '\n';
3490 options->break_opt = -1;
3491 options->rename_limit = -1;
3492 options->dirstat_permille = diff_dirstat_permille_default;
3493 options->context = diff_context_default;
3494 options->interhunkcontext = diff_interhunk_context_default;
3495 options->ws_error_highlight = ws_error_highlight_default;
3496 DIFF_OPT_SET(options, RENAME_EMPTY);
3497
3498 /* pathchange left =NULL by default */
3499 options->change = diff_change;
3500 options->add_remove = diff_addremove;
3501 options->use_color = diff_use_color_default;
3502 options->detect_rename = diff_detect_rename_default;
3503 options->xdl_opts |= diff_algorithm;
3504 if (diff_indent_heuristic)
3505 DIFF_XDL_SET(options, INDENT_HEURISTIC);
3506
3507 options->orderfile = diff_order_file_cfg;
3508
3509 if (diff_no_prefix) {
3510 options->a_prefix = options->b_prefix = "";
3511 } else if (!diff_mnemonic_prefix) {
3512 options->a_prefix = "a/";
3513 options->b_prefix = "b/";
3514 }
3515}
3516
3517void diff_setup_done(struct diff_options *options)
3518{
3519 int count = 0;
3520
3521 if (options->set_default)
3522 options->set_default(options);
3523
3524 if (options->output_format & DIFF_FORMAT_NAME)
3525 count++;
3526 if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3527 count++;
3528 if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3529 count++;
3530 if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3531 count++;
3532 if (count > 1)
3533 die(_("--name-only, --name-status, --check and -s are mutually exclusive"));
3534
3535 /*
3536 * Most of the time we can say "there are changes"
3537 * only by checking if there are changed paths, but
3538 * --ignore-whitespace* options force us to look
3539 * inside contents.
3540 */
3541
3542 if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3543 DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3544 DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3545 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3546 else
3547 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3548
3549 if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3550 options->detect_rename = DIFF_DETECT_COPY;
3551
3552 if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3553 options->prefix = NULL;
3554 if (options->prefix)
3555 options->prefix_length = strlen(options->prefix);
3556 else
3557 options->prefix_length = 0;
3558
3559 if (options->output_format & (DIFF_FORMAT_NAME |
3560 DIFF_FORMAT_NAME_STATUS |
3561 DIFF_FORMAT_CHECKDIFF |
3562 DIFF_FORMAT_NO_OUTPUT))
3563 options->output_format &= ~(DIFF_FORMAT_RAW |
3564 DIFF_FORMAT_NUMSTAT |
3565 DIFF_FORMAT_DIFFSTAT |
3566 DIFF_FORMAT_SHORTSTAT |
3567 DIFF_FORMAT_DIRSTAT |
3568 DIFF_FORMAT_SUMMARY |
3569 DIFF_FORMAT_PATCH);
3570
3571 /*
3572 * These cases always need recursive; we do not drop caller-supplied
3573 * recursive bits for other formats here.
3574 */
3575 if (options->output_format & (DIFF_FORMAT_PATCH |
3576 DIFF_FORMAT_NUMSTAT |
3577 DIFF_FORMAT_DIFFSTAT |
3578 DIFF_FORMAT_SHORTSTAT |
3579 DIFF_FORMAT_DIRSTAT |
3580 DIFF_FORMAT_SUMMARY |
3581 DIFF_FORMAT_CHECKDIFF))
3582 DIFF_OPT_SET(options, RECURSIVE);
3583 /*
3584 * Also pickaxe would not work very well if you do not say recursive
3585 */
3586 if (options->pickaxe)
3587 DIFF_OPT_SET(options, RECURSIVE);
3588 /*
3589 * When patches are generated, submodules diffed against the work tree
3590 * must be checked for dirtiness too so it can be shown in the output
3591 */
3592 if (options->output_format & DIFF_FORMAT_PATCH)
3593 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3594
3595 if (options->detect_rename && options->rename_limit < 0)
3596 options->rename_limit = diff_rename_limit_default;
3597 if (options->setup & DIFF_SETUP_USE_CACHE) {
3598 if (!active_cache)
3599 /* read-cache does not die even when it fails
3600 * so it is safe for us to do this here. Also
3601 * it does not smudge active_cache or active_nr
3602 * when it fails, so we do not have to worry about
3603 * cleaning it up ourselves either.
3604 */
3605 read_cache();
3606 }
3607 if (40 < options->abbrev)
3608 options->abbrev = 40; /* full */
3609
3610 /*
3611 * It does not make sense to show the first hit we happened
3612 * to have found. It does not make sense not to return with
3613 * exit code in such a case either.
3614 */
3615 if (DIFF_OPT_TST(options, QUICK)) {
3616 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3617 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3618 }
3619
3620 options->diff_path_counter = 0;
3621
3622 if (DIFF_OPT_TST(options, FOLLOW_RENAMES) && options->pathspec.nr != 1)
3623 die(_("--follow requires exactly one pathspec"));
3624}
3625
3626static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3627{
3628 char c, *eq;
3629 int len;
3630
3631 if (*arg != '-')
3632 return 0;
3633 c = *++arg;
3634 if (!c)
3635 return 0;
3636 if (c == arg_short) {
3637 c = *++arg;
3638 if (!c)
3639 return 1;
3640 if (val && isdigit(c)) {
3641 char *end;
3642 int n = strtoul(arg, &end, 10);
3643 if (*end)
3644 return 0;
3645 *val = n;
3646 return 1;
3647 }
3648 return 0;
3649 }
3650 if (c != '-')
3651 return 0;
3652 arg++;
3653 eq = strchrnul(arg, '=');
3654 len = eq - arg;
3655 if (!len || strncmp(arg, arg_long, len))
3656 return 0;
3657 if (*eq) {
3658 int n;
3659 char *end;
3660 if (!isdigit(*++eq))
3661 return 0;
3662 n = strtoul(eq, &end, 10);
3663 if (*end)
3664 return 0;
3665 *val = n;
3666 }
3667 return 1;
3668}
3669
3670static int diff_scoreopt_parse(const char *opt);
3671
3672static inline int short_opt(char opt, const char **argv,
3673 const char **optarg)
3674{
3675 const char *arg = argv[0];
3676 if (arg[0] != '-' || arg[1] != opt)
3677 return 0;
3678 if (arg[2] != '\0') {
3679 *optarg = arg + 2;
3680 return 1;
3681 }
3682 if (!argv[1])
3683 die("Option '%c' requires a value", opt);
3684 *optarg = argv[1];
3685 return 2;
3686}
3687
3688int parse_long_opt(const char *opt, const char **argv,
3689 const char **optarg)
3690{
3691 const char *arg = argv[0];
3692 if (!skip_prefix(arg, "--", &arg))
3693 return 0;
3694 if (!skip_prefix(arg, opt, &arg))
3695 return 0;
3696 if (*arg == '=') { /* stuck form: --option=value */
3697 *optarg = arg + 1;
3698 return 1;
3699 }
3700 if (*arg != '\0')
3701 return 0;
3702 /* separate form: --option value */
3703 if (!argv[1])
3704 die("Option '--%s' requires a value", opt);
3705 *optarg = argv[1];
3706 return 2;
3707}
3708
3709static int stat_opt(struct diff_options *options, const char **av)
3710{
3711 const char *arg = av[0];
3712 char *end;
3713 int width = options->stat_width;
3714 int name_width = options->stat_name_width;
3715 int graph_width = options->stat_graph_width;
3716 int count = options->stat_count;
3717 int argcount = 1;
3718
3719 if (!skip_prefix(arg, "--stat", &arg))
3720 die("BUG: stat option does not begin with --stat: %s", arg);
3721 end = (char *)arg;
3722
3723 switch (*arg) {
3724 case '-':
3725 if (skip_prefix(arg, "-width", &arg)) {
3726 if (*arg == '=')
3727 width = strtoul(arg + 1, &end, 10);
3728 else if (!*arg && !av[1])
3729 die_want_option("--stat-width");
3730 else if (!*arg) {
3731 width = strtoul(av[1], &end, 10);
3732 argcount = 2;
3733 }
3734 } else if (skip_prefix(arg, "-name-width", &arg)) {
3735 if (*arg == '=')
3736 name_width = strtoul(arg + 1, &end, 10);
3737 else if (!*arg && !av[1])
3738 die_want_option("--stat-name-width");
3739 else if (!*arg) {
3740 name_width = strtoul(av[1], &end, 10);
3741 argcount = 2;
3742 }
3743 } else if (skip_prefix(arg, "-graph-width", &arg)) {
3744 if (*arg == '=')
3745 graph_width = strtoul(arg + 1, &end, 10);
3746 else if (!*arg && !av[1])
3747 die_want_option("--stat-graph-width");
3748 else if (!*arg) {
3749 graph_width = strtoul(av[1], &end, 10);
3750 argcount = 2;
3751 }
3752 } else if (skip_prefix(arg, "-count", &arg)) {
3753 if (*arg == '=')
3754 count = strtoul(arg + 1, &end, 10);
3755 else if (!*arg && !av[1])
3756 die_want_option("--stat-count");
3757 else if (!*arg) {
3758 count = strtoul(av[1], &end, 10);
3759 argcount = 2;
3760 }
3761 }
3762 break;
3763 case '=':
3764 width = strtoul(arg+1, &end, 10);
3765 if (*end == ',')
3766 name_width = strtoul(end+1, &end, 10);
3767 if (*end == ',')
3768 count = strtoul(end+1, &end, 10);
3769 }
3770
3771 /* Important! This checks all the error cases! */
3772 if (*end)
3773 return 0;
3774 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3775 options->stat_name_width = name_width;
3776 options->stat_graph_width = graph_width;
3777 options->stat_width = width;
3778 options->stat_count = count;
3779 return argcount;
3780}
3781
3782static int parse_dirstat_opt(struct diff_options *options, const char *params)
3783{
3784 struct strbuf errmsg = STRBUF_INIT;
3785 if (parse_dirstat_params(options, params, &errmsg))
3786 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3787 errmsg.buf);
3788 strbuf_release(&errmsg);
3789 /*
3790 * The caller knows a dirstat-related option is given from the command
3791 * line; allow it to say "return this_function();"
3792 */
3793 options->output_format |= DIFF_FORMAT_DIRSTAT;
3794 return 1;
3795}
3796
3797static int parse_submodule_opt(struct diff_options *options, const char *value)
3798{
3799 if (parse_submodule_params(options, value))
3800 die(_("Failed to parse --submodule option parameter: '%s'"),
3801 value);
3802 return 1;
3803}
3804
3805static const char diff_status_letters[] = {
3806 DIFF_STATUS_ADDED,
3807 DIFF_STATUS_COPIED,
3808 DIFF_STATUS_DELETED,
3809 DIFF_STATUS_MODIFIED,
3810 DIFF_STATUS_RENAMED,
3811 DIFF_STATUS_TYPE_CHANGED,
3812 DIFF_STATUS_UNKNOWN,
3813 DIFF_STATUS_UNMERGED,
3814 DIFF_STATUS_FILTER_AON,
3815 DIFF_STATUS_FILTER_BROKEN,
3816 '\0',
3817};
3818
3819static unsigned int filter_bit['Z' + 1];
3820
3821static void prepare_filter_bits(void)
3822{
3823 int i;
3824
3825 if (!filter_bit[DIFF_STATUS_ADDED]) {
3826 for (i = 0; diff_status_letters[i]; i++)
3827 filter_bit[(int) diff_status_letters[i]] = (1 << i);
3828 }
3829}
3830
3831static unsigned filter_bit_tst(char status, const struct diff_options *opt)
3832{
3833 return opt->filter & filter_bit[(int) status];
3834}
3835
3836static int parse_diff_filter_opt(const char *optarg, struct diff_options *opt)
3837{
3838 int i, optch;
3839
3840 prepare_filter_bits();
3841
3842 /*
3843 * If there is a negation e.g. 'd' in the input, and we haven't
3844 * initialized the filter field with another --diff-filter, start
3845 * from full set of bits, except for AON.
3846 */
3847 if (!opt->filter) {
3848 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3849 if (optch < 'a' || 'z' < optch)
3850 continue;
3851 opt->filter = (1 << (ARRAY_SIZE(diff_status_letters) - 1)) - 1;
3852 opt->filter &= ~filter_bit[DIFF_STATUS_FILTER_AON];
3853 break;
3854 }
3855 }
3856
3857 for (i = 0; (optch = optarg[i]) != '\0'; i++) {
3858 unsigned int bit;
3859 int negate;
3860
3861 if ('a' <= optch && optch <= 'z') {
3862 negate = 1;
3863 optch = toupper(optch);
3864 } else {
3865 negate = 0;
3866 }
3867
3868 bit = (0 <= optch && optch <= 'Z') ? filter_bit[optch] : 0;
3869 if (!bit)
3870 return optarg[i];
3871 if (negate)
3872 opt->filter &= ~bit;
3873 else
3874 opt->filter |= bit;
3875 }
3876 return 0;
3877}
3878
3879static void enable_patch_output(int *fmt) {
3880 *fmt &= ~DIFF_FORMAT_NO_OUTPUT;
3881 *fmt |= DIFF_FORMAT_PATCH;
3882}
3883
3884static int parse_ws_error_highlight_opt(struct diff_options *opt, const char *arg)
3885{
3886 int val = parse_ws_error_highlight(arg);
3887
3888 if (val < 0) {
3889 error("unknown value after ws-error-highlight=%.*s",
3890 -1 - val, arg);
3891 return 0;
3892 }
3893 opt->ws_error_highlight = val;
3894 return 1;
3895}
3896
3897int diff_opt_parse(struct diff_options *options,
3898 const char **av, int ac, const char *prefix)
3899{
3900 const char *arg = av[0];
3901 const char *optarg;
3902 int argcount;
3903
3904 if (!prefix)
3905 prefix = "";
3906
3907 /* Output format options */
3908 if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch")
3909 || opt_arg(arg, 'U', "unified", &options->context))
3910 enable_patch_output(&options->output_format);
3911 else if (!strcmp(arg, "--raw"))
3912 options->output_format |= DIFF_FORMAT_RAW;
3913 else if (!strcmp(arg, "--patch-with-raw")) {
3914 enable_patch_output(&options->output_format);
3915 options->output_format |= DIFF_FORMAT_RAW;
3916 } else if (!strcmp(arg, "--numstat"))
3917 options->output_format |= DIFF_FORMAT_NUMSTAT;
3918 else if (!strcmp(arg, "--shortstat"))
3919 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3920 else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3921 return parse_dirstat_opt(options, "");
3922 else if (skip_prefix(arg, "-X", &arg))
3923 return parse_dirstat_opt(options, arg);
3924 else if (skip_prefix(arg, "--dirstat=", &arg))
3925 return parse_dirstat_opt(options, arg);
3926 else if (!strcmp(arg, "--cumulative"))
3927 return parse_dirstat_opt(options, "cumulative");
3928 else if (!strcmp(arg, "--dirstat-by-file"))
3929 return parse_dirstat_opt(options, "files");
3930 else if (skip_prefix(arg, "--dirstat-by-file=", &arg)) {
3931 parse_dirstat_opt(options, "files");
3932 return parse_dirstat_opt(options, arg);
3933 }
3934 else if (!strcmp(arg, "--check"))
3935 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3936 else if (!strcmp(arg, "--summary"))
3937 options->output_format |= DIFF_FORMAT_SUMMARY;
3938 else if (!strcmp(arg, "--patch-with-stat")) {
3939 enable_patch_output(&options->output_format);
3940 options->output_format |= DIFF_FORMAT_DIFFSTAT;
3941 } else if (!strcmp(arg, "--name-only"))
3942 options->output_format |= DIFF_FORMAT_NAME;
3943 else if (!strcmp(arg, "--name-status"))
3944 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3945 else if (!strcmp(arg, "-s") || !strcmp(arg, "--no-patch"))
3946 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3947 else if (starts_with(arg, "--stat"))
3948 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3949 return stat_opt(options, av);
3950
3951 /* renames options */
3952 else if (starts_with(arg, "-B") || starts_with(arg, "--break-rewrites=") ||
3953 !strcmp(arg, "--break-rewrites")) {
3954 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3955 return error("invalid argument to -B: %s", arg+2);
3956 }
3957 else if (starts_with(arg, "-M") || starts_with(arg, "--find-renames=") ||
3958 !strcmp(arg, "--find-renames")) {
3959 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3960 return error("invalid argument to -M: %s", arg+2);
3961 options->detect_rename = DIFF_DETECT_RENAME;
3962 }
3963 else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3964 options->irreversible_delete = 1;
3965 }
3966 else if (starts_with(arg, "-C") || starts_with(arg, "--find-copies=") ||
3967 !strcmp(arg, "--find-copies")) {
3968 if (options->detect_rename == DIFF_DETECT_COPY)
3969 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3970 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3971 return error("invalid argument to -C: %s", arg+2);
3972 options->detect_rename = DIFF_DETECT_COPY;
3973 }
3974 else if (!strcmp(arg, "--no-renames"))
3975 options->detect_rename = 0;
3976 else if (!strcmp(arg, "--rename-empty"))
3977 DIFF_OPT_SET(options, RENAME_EMPTY);
3978 else if (!strcmp(arg, "--no-rename-empty"))
3979 DIFF_OPT_CLR(options, RENAME_EMPTY);
3980 else if (!strcmp(arg, "--relative"))
3981 DIFF_OPT_SET(options, RELATIVE_NAME);
3982 else if (skip_prefix(arg, "--relative=", &arg)) {
3983 DIFF_OPT_SET(options, RELATIVE_NAME);
3984 options->prefix = arg;
3985 }
3986
3987 /* xdiff options */
3988 else if (!strcmp(arg, "--minimal"))
3989 DIFF_XDL_SET(options, NEED_MINIMAL);
3990 else if (!strcmp(arg, "--no-minimal"))
3991 DIFF_XDL_CLR(options, NEED_MINIMAL);
3992 else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3993 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3994 else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3995 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3996 else if (!strcmp(arg, "--ignore-space-at-eol"))
3997 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3998 else if (!strcmp(arg, "--ignore-blank-lines"))
3999 DIFF_XDL_SET(options, IGNORE_BLANK_LINES);
4000 else if (!strcmp(arg, "--indent-heuristic"))
4001 DIFF_XDL_SET(options, INDENT_HEURISTIC);
4002 else if (!strcmp(arg, "--no-indent-heuristic"))
4003 DIFF_XDL_CLR(options, INDENT_HEURISTIC);
4004 else if (!strcmp(arg, "--patience"))
4005 options->xdl_opts = DIFF_WITH_ALG(options, PATIENCE_DIFF);
4006 else if (!strcmp(arg, "--histogram"))
4007 options->xdl_opts = DIFF_WITH_ALG(options, HISTOGRAM_DIFF);
4008 else if ((argcount = parse_long_opt("diff-algorithm", av, &optarg))) {
4009 long value = parse_algorithm_value(optarg);
4010 if (value < 0)
4011 return error("option diff-algorithm accepts \"myers\", "
4012 "\"minimal\", \"patience\" and \"histogram\"");
4013 /* clear out previous settings */
4014 DIFF_XDL_CLR(options, NEED_MINIMAL);
4015 options->xdl_opts &= ~XDF_DIFF_ALGORITHM_MASK;
4016 options->xdl_opts |= value;
4017 return argcount;
4018 }
4019
4020 /* flags options */
4021 else if (!strcmp(arg, "--binary")) {
4022 enable_patch_output(&options->output_format);
4023 DIFF_OPT_SET(options, BINARY);
4024 }
4025 else if (!strcmp(arg, "--full-index"))
4026 DIFF_OPT_SET(options, FULL_INDEX);
4027 else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
4028 DIFF_OPT_SET(options, TEXT);
4029 else if (!strcmp(arg, "-R"))
4030 DIFF_OPT_SET(options, REVERSE_DIFF);
4031 else if (!strcmp(arg, "--find-copies-harder"))
4032 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
4033 else if (!strcmp(arg, "--follow"))
4034 DIFF_OPT_SET(options, FOLLOW_RENAMES);
4035 else if (!strcmp(arg, "--no-follow")) {
4036 DIFF_OPT_CLR(options, FOLLOW_RENAMES);
4037 DIFF_OPT_CLR(options, DEFAULT_FOLLOW_RENAMES);
4038 } else if (!strcmp(arg, "--color"))
4039 options->use_color = 1;
4040 else if (skip_prefix(arg, "--color=", &arg)) {
4041 int value = git_config_colorbool(NULL, arg);
4042 if (value < 0)
4043 return error("option `color' expects \"always\", \"auto\", or \"never\"");
4044 options->use_color = value;
4045 }
4046 else if (!strcmp(arg, "--no-color"))
4047 options->use_color = 0;
4048 else if (!strcmp(arg, "--color-words")) {
4049 options->use_color = 1;
4050 options->word_diff = DIFF_WORDS_COLOR;
4051 }
4052 else if (skip_prefix(arg, "--color-words=", &arg)) {
4053 options->use_color = 1;
4054 options->word_diff = DIFF_WORDS_COLOR;
4055 options->word_regex = arg;
4056 }
4057 else if (!strcmp(arg, "--word-diff")) {
4058 if (options->word_diff == DIFF_WORDS_NONE)
4059 options->word_diff = DIFF_WORDS_PLAIN;
4060 }
4061 else if (skip_prefix(arg, "--word-diff=", &arg)) {
4062 if (!strcmp(arg, "plain"))
4063 options->word_diff = DIFF_WORDS_PLAIN;
4064 else if (!strcmp(arg, "color")) {
4065 options->use_color = 1;
4066 options->word_diff = DIFF_WORDS_COLOR;
4067 }
4068 else if (!strcmp(arg, "porcelain"))
4069 options->word_diff = DIFF_WORDS_PORCELAIN;
4070 else if (!strcmp(arg, "none"))
4071 options->word_diff = DIFF_WORDS_NONE;
4072 else
4073 die("bad --word-diff argument: %s", arg);
4074 }
4075 else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
4076 if (options->word_diff == DIFF_WORDS_NONE)
4077 options->word_diff = DIFF_WORDS_PLAIN;
4078 options->word_regex = optarg;
4079 return argcount;
4080 }
4081 else if (!strcmp(arg, "--exit-code"))
4082 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
4083 else if (!strcmp(arg, "--quiet"))
4084 DIFF_OPT_SET(options, QUICK);
4085 else if (!strcmp(arg, "--ext-diff"))
4086 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
4087 else if (!strcmp(arg, "--no-ext-diff"))
4088 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
4089 else if (!strcmp(arg, "--textconv"))
4090 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
4091 else if (!strcmp(arg, "--no-textconv"))
4092 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
4093 else if (!strcmp(arg, "--ignore-submodules")) {
4094 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
4095 handle_ignore_submodules_arg(options, "all");
4096 } else if (skip_prefix(arg, "--ignore-submodules=", &arg)) {
4097 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
4098 handle_ignore_submodules_arg(options, arg);
4099 } else if (!strcmp(arg, "--submodule"))
4100 options->submodule_format = DIFF_SUBMODULE_LOG;
4101 else if (skip_prefix(arg, "--submodule=", &arg))
4102 return parse_submodule_opt(options, arg);
4103 else if (skip_prefix(arg, "--ws-error-highlight=", &arg))
4104 return parse_ws_error_highlight_opt(options, arg);
4105 else if (!strcmp(arg, "--ita-invisible-in-index"))
4106 options->ita_invisible_in_index = 1;
4107 else if (!strcmp(arg, "--ita-visible-in-index"))
4108 options->ita_invisible_in_index = 0;
4109
4110 /* misc options */
4111 else if (!strcmp(arg, "-z"))
4112 options->line_termination = 0;
4113 else if ((argcount = short_opt('l', av, &optarg))) {
4114 options->rename_limit = strtoul(optarg, NULL, 10);
4115 return argcount;
4116 }
4117 else if ((argcount = short_opt('S', av, &optarg))) {
4118 options->pickaxe = optarg;
4119 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
4120 return argcount;
4121 } else if ((argcount = short_opt('G', av, &optarg))) {
4122 options->pickaxe = optarg;
4123 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
4124 return argcount;
4125 }
4126 else if (!strcmp(arg, "--pickaxe-all"))
4127 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
4128 else if (!strcmp(arg, "--pickaxe-regex"))
4129 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
4130 else if ((argcount = short_opt('O', av, &optarg))) {
4131 options->orderfile = prefix_filename(prefix, optarg);
4132 return argcount;
4133 }
4134 else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
4135 int offending = parse_diff_filter_opt(optarg, options);
4136 if (offending)
4137 die("unknown change class '%c' in --diff-filter=%s",
4138 offending, optarg);
4139 return argcount;
4140 }
4141 else if (!strcmp(arg, "--no-abbrev"))
4142 options->abbrev = 0;
4143 else if (!strcmp(arg, "--abbrev"))
4144 options->abbrev = DEFAULT_ABBREV;
4145 else if (skip_prefix(arg, "--abbrev=", &arg)) {
4146 options->abbrev = strtoul(arg, NULL, 10);
4147 if (options->abbrev < MINIMUM_ABBREV)
4148 options->abbrev = MINIMUM_ABBREV;
4149 else if (40 < options->abbrev)
4150 options->abbrev = 40;
4151 }
4152 else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
4153 options->a_prefix = optarg;
4154 return argcount;
4155 }
4156 else if ((argcount = parse_long_opt("line-prefix", av, &optarg))) {
4157 options->line_prefix = optarg;
4158 options->line_prefix_length = strlen(options->line_prefix);
4159 graph_setup_line_prefix(options);
4160 return argcount;
4161 }
4162 else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
4163 options->b_prefix = optarg;
4164 return argcount;
4165 }
4166 else if (!strcmp(arg, "--no-prefix"))
4167 options->a_prefix = options->b_prefix = "";
4168 else if (opt_arg(arg, '\0', "inter-hunk-context",
4169 &options->interhunkcontext))
4170 ;
4171 else if (!strcmp(arg, "-W"))
4172 DIFF_OPT_SET(options, FUNCCONTEXT);
4173 else if (!strcmp(arg, "--function-context"))
4174 DIFF_OPT_SET(options, FUNCCONTEXT);
4175 else if (!strcmp(arg, "--no-function-context"))
4176 DIFF_OPT_CLR(options, FUNCCONTEXT);
4177 else if ((argcount = parse_long_opt("output", av, &optarg))) {
4178 char *path = prefix_filename(prefix, optarg);
4179 options->file = xfopen(path, "w");
4180 options->close_file = 1;
4181 if (options->use_color != GIT_COLOR_ALWAYS)
4182 options->use_color = GIT_COLOR_NEVER;
4183 free(path);
4184 return argcount;
4185 } else
4186 return 0;
4187 return 1;
4188}
4189
4190int parse_rename_score(const char **cp_p)
4191{
4192 unsigned long num, scale;
4193 int ch, dot;
4194 const char *cp = *cp_p;
4195
4196 num = 0;
4197 scale = 1;
4198 dot = 0;
4199 for (;;) {
4200 ch = *cp;
4201 if ( !dot && ch == '.' ) {
4202 scale = 1;
4203 dot = 1;
4204 } else if ( ch == '%' ) {
4205 scale = dot ? scale*100 : 100;
4206 cp++; /* % is always at the end */
4207 break;
4208 } else if ( ch >= '0' && ch <= '9' ) {
4209 if ( scale < 100000 ) {
4210 scale *= 10;
4211 num = (num*10) + (ch-'0');
4212 }
4213 } else {
4214 break;
4215 }
4216 cp++;
4217 }
4218 *cp_p = cp;
4219
4220 /* user says num divided by scale and we say internally that
4221 * is MAX_SCORE * num / scale.
4222 */
4223 return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
4224}
4225
4226static int diff_scoreopt_parse(const char *opt)
4227{
4228 int opt1, opt2, cmd;
4229
4230 if (*opt++ != '-')
4231 return -1;
4232 cmd = *opt++;
4233 if (cmd == '-') {
4234 /* convert the long-form arguments into short-form versions */
4235 if (skip_prefix(opt, "break-rewrites", &opt)) {
4236 if (*opt == 0 || *opt++ == '=')
4237 cmd = 'B';
4238 } else if (skip_prefix(opt, "find-copies", &opt)) {
4239 if (*opt == 0 || *opt++ == '=')
4240 cmd = 'C';
4241 } else if (skip_prefix(opt, "find-renames", &opt)) {
4242 if (*opt == 0 || *opt++ == '=')
4243 cmd = 'M';
4244 }
4245 }
4246 if (cmd != 'M' && cmd != 'C' && cmd != 'B')
4247 return -1; /* that is not a -M, -C, or -B option */
4248
4249 opt1 = parse_rename_score(&opt);
4250 if (cmd != 'B')
4251 opt2 = 0;
4252 else {
4253 if (*opt == 0)
4254 opt2 = 0;
4255 else if (*opt != '/')
4256 return -1; /* we expect -B80/99 or -B80 */
4257 else {
4258 opt++;
4259 opt2 = parse_rename_score(&opt);
4260 }
4261 }
4262 if (*opt != 0)
4263 return -1;
4264 return opt1 | (opt2 << 16);
4265}
4266
4267struct diff_queue_struct diff_queued_diff;
4268
4269void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
4270{
4271 ALLOC_GROW(queue->queue, queue->nr + 1, queue->alloc);
4272 queue->queue[queue->nr++] = dp;
4273}
4274
4275struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
4276 struct diff_filespec *one,
4277 struct diff_filespec *two)
4278{
4279 struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
4280 dp->one = one;
4281 dp->two = two;
4282 if (queue)
4283 diff_q(queue, dp);
4284 return dp;
4285}
4286
4287void diff_free_filepair(struct diff_filepair *p)
4288{
4289 free_filespec(p->one);
4290 free_filespec(p->two);
4291 free(p);
4292}
4293
4294const char *diff_aligned_abbrev(const struct object_id *oid, int len)
4295{
4296 int abblen;
4297 const char *abbrev;
4298
4299 if (len == GIT_SHA1_HEXSZ)
4300 return oid_to_hex(oid);
4301
4302 abbrev = diff_abbrev_oid(oid, len);
4303 abblen = strlen(abbrev);
4304
4305 /*
4306 * In well-behaved cases, where the abbbreviated result is the
4307 * same as the requested length, append three dots after the
4308 * abbreviation (hence the whole logic is limited to the case
4309 * where abblen < 37); when the actual abbreviated result is a
4310 * bit longer than the requested length, we reduce the number
4311 * of dots so that they match the well-behaved ones. However,
4312 * if the actual abbreviation is longer than the requested
4313 * length by more than three, we give up on aligning, and add
4314 * three dots anyway, to indicate that the output is not the
4315 * full object name. Yes, this may be suboptimal, but this
4316 * appears only in "diff --raw --abbrev" output and it is not
4317 * worth the effort to change it now. Note that this would
4318 * likely to work fine when the automatic sizing of default
4319 * abbreviation length is used--we would be fed -1 in "len" in
4320 * that case, and will end up always appending three-dots, but
4321 * the automatic sizing is supposed to give abblen that ensures
4322 * uniqueness across all objects (statistically speaking).
4323 */
4324 if (abblen < GIT_SHA1_HEXSZ - 3) {
4325 static char hex[GIT_MAX_HEXSZ + 1];
4326 if (len < abblen && abblen <= len + 2)
4327 xsnprintf(hex, sizeof(hex), "%s%.*s", abbrev, len+3-abblen, "..");
4328 else
4329 xsnprintf(hex, sizeof(hex), "%s...", abbrev);
4330 return hex;
4331 }
4332
4333 return oid_to_hex(oid);
4334}
4335
4336static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
4337{
4338 int line_termination = opt->line_termination;
4339 int inter_name_termination = line_termination ? '\t' : '\0';
4340
4341 fprintf(opt->file, "%s", diff_line_prefix(opt));
4342 if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
4343 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
4344 diff_aligned_abbrev(&p->one->oid, opt->abbrev));
4345 fprintf(opt->file, "%s ",
4346 diff_aligned_abbrev(&p->two->oid, opt->abbrev));
4347 }
4348 if (p->score) {
4349 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
4350 inter_name_termination);
4351 } else {
4352 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
4353 }
4354
4355 if (p->status == DIFF_STATUS_COPIED ||
4356 p->status == DIFF_STATUS_RENAMED) {
4357 const char *name_a, *name_b;
4358 name_a = p->one->path;
4359 name_b = p->two->path;
4360 strip_prefix(opt->prefix_length, &name_a, &name_b);
4361 write_name_quoted(name_a, opt->file, inter_name_termination);
4362 write_name_quoted(name_b, opt->file, line_termination);
4363 } else {
4364 const char *name_a, *name_b;
4365 name_a = p->one->mode ? p->one->path : p->two->path;
4366 name_b = NULL;
4367 strip_prefix(opt->prefix_length, &name_a, &name_b);
4368 write_name_quoted(name_a, opt->file, line_termination);
4369 }
4370}
4371
4372int diff_unmodified_pair(struct diff_filepair *p)
4373{
4374 /* This function is written stricter than necessary to support
4375 * the currently implemented transformers, but the idea is to
4376 * let transformers to produce diff_filepairs any way they want,
4377 * and filter and clean them up here before producing the output.
4378 */
4379 struct diff_filespec *one = p->one, *two = p->two;
4380
4381 if (DIFF_PAIR_UNMERGED(p))
4382 return 0; /* unmerged is interesting */
4383
4384 /* deletion, addition, mode or type change
4385 * and rename are all interesting.
4386 */
4387 if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
4388 DIFF_PAIR_MODE_CHANGED(p) ||
4389 strcmp(one->path, two->path))
4390 return 0;
4391
4392 /* both are valid and point at the same path. that is, we are
4393 * dealing with a change.
4394 */
4395 if (one->oid_valid && two->oid_valid &&
4396 !oidcmp(&one->oid, &two->oid) &&
4397 !one->dirty_submodule && !two->dirty_submodule)
4398 return 1; /* no change */
4399 if (!one->oid_valid && !two->oid_valid)
4400 return 1; /* both look at the same file on the filesystem. */
4401 return 0;
4402}
4403
4404static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
4405{
4406 if (diff_unmodified_pair(p))
4407 return;
4408
4409 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4410 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4411 return; /* no tree diffs in patch format */
4412
4413 run_diff(p, o);
4414}
4415
4416static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
4417 struct diffstat_t *diffstat)
4418{
4419 if (diff_unmodified_pair(p))
4420 return;
4421
4422 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4423 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4424 return; /* no useful stat for tree diffs */
4425
4426 run_diffstat(p, o, diffstat);
4427}
4428
4429static void diff_flush_checkdiff(struct diff_filepair *p,
4430 struct diff_options *o)
4431{
4432 if (diff_unmodified_pair(p))
4433 return;
4434
4435 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4436 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4437 return; /* nothing to check in tree diffs */
4438
4439 run_checkdiff(p, o);
4440}
4441
4442int diff_queue_is_empty(void)
4443{
4444 struct diff_queue_struct *q = &diff_queued_diff;
4445 int i;
4446 for (i = 0; i < q->nr; i++)
4447 if (!diff_unmodified_pair(q->queue[i]))
4448 return 0;
4449 return 1;
4450}
4451
4452#if DIFF_DEBUG
4453void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
4454{
4455 fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
4456 x, one ? one : "",
4457 s->path,
4458 DIFF_FILE_VALID(s) ? "valid" : "invalid",
4459 s->mode,
4460 s->oid_valid ? oid_to_hex(&s->oid) : "");
4461 fprintf(stderr, "queue[%d] %s size %lu\n",
4462 x, one ? one : "",
4463 s->size);
4464}
4465
4466void diff_debug_filepair(const struct diff_filepair *p, int i)
4467{
4468 diff_debug_filespec(p->one, i, "one");
4469 diff_debug_filespec(p->two, i, "two");
4470 fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
4471 p->score, p->status ? p->status : '?',
4472 p->one->rename_used, p->broken_pair);
4473}
4474
4475void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
4476{
4477 int i;
4478 if (msg)
4479 fprintf(stderr, "%s\n", msg);
4480 fprintf(stderr, "q->nr = %d\n", q->nr);
4481 for (i = 0; i < q->nr; i++) {
4482 struct diff_filepair *p = q->queue[i];
4483 diff_debug_filepair(p, i);
4484 }
4485}
4486#endif
4487
4488static void diff_resolve_rename_copy(void)
4489{
4490 int i;
4491 struct diff_filepair *p;
4492 struct diff_queue_struct *q = &diff_queued_diff;
4493
4494 diff_debug_queue("resolve-rename-copy", q);
4495
4496 for (i = 0; i < q->nr; i++) {
4497 p = q->queue[i];
4498 p->status = 0; /* undecided */
4499 if (DIFF_PAIR_UNMERGED(p))
4500 p->status = DIFF_STATUS_UNMERGED;
4501 else if (!DIFF_FILE_VALID(p->one))
4502 p->status = DIFF_STATUS_ADDED;
4503 else if (!DIFF_FILE_VALID(p->two))
4504 p->status = DIFF_STATUS_DELETED;
4505 else if (DIFF_PAIR_TYPE_CHANGED(p))
4506 p->status = DIFF_STATUS_TYPE_CHANGED;
4507
4508 /* from this point on, we are dealing with a pair
4509 * whose both sides are valid and of the same type, i.e.
4510 * either in-place edit or rename/copy edit.
4511 */
4512 else if (DIFF_PAIR_RENAME(p)) {
4513 /*
4514 * A rename might have re-connected a broken
4515 * pair up, causing the pathnames to be the
4516 * same again. If so, that's not a rename at
4517 * all, just a modification..
4518 *
4519 * Otherwise, see if this source was used for
4520 * multiple renames, in which case we decrement
4521 * the count, and call it a copy.
4522 */
4523 if (!strcmp(p->one->path, p->two->path))
4524 p->status = DIFF_STATUS_MODIFIED;
4525 else if (--p->one->rename_used > 0)
4526 p->status = DIFF_STATUS_COPIED;
4527 else
4528 p->status = DIFF_STATUS_RENAMED;
4529 }
4530 else if (oidcmp(&p->one->oid, &p->two->oid) ||
4531 p->one->mode != p->two->mode ||
4532 p->one->dirty_submodule ||
4533 p->two->dirty_submodule ||
4534 is_null_oid(&p->one->oid))
4535 p->status = DIFF_STATUS_MODIFIED;
4536 else {
4537 /* This is a "no-change" entry and should not
4538 * happen anymore, but prepare for broken callers.
4539 */
4540 error("feeding unmodified %s to diffcore",
4541 p->one->path);
4542 p->status = DIFF_STATUS_UNKNOWN;
4543 }
4544 }
4545 diff_debug_queue("resolve-rename-copy done", q);
4546}
4547
4548static int check_pair_status(struct diff_filepair *p)
4549{
4550 switch (p->status) {
4551 case DIFF_STATUS_UNKNOWN:
4552 return 0;
4553 case 0:
4554 die("internal error in diff-resolve-rename-copy");
4555 default:
4556 return 1;
4557 }
4558}
4559
4560static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
4561{
4562 int fmt = opt->output_format;
4563
4564 if (fmt & DIFF_FORMAT_CHECKDIFF)
4565 diff_flush_checkdiff(p, opt);
4566 else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
4567 diff_flush_raw(p, opt);
4568 else if (fmt & DIFF_FORMAT_NAME) {
4569 const char *name_a, *name_b;
4570 name_a = p->two->path;
4571 name_b = NULL;
4572 strip_prefix(opt->prefix_length, &name_a, &name_b);
4573 fprintf(opt->file, "%s", diff_line_prefix(opt));
4574 write_name_quoted(name_a, opt->file, opt->line_termination);
4575 }
4576}
4577
4578static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
4579{
4580 if (fs->mode)
4581 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
4582 else
4583 fprintf(file, " %s ", newdelete);
4584 write_name_quoted(fs->path, file, '\n');
4585}
4586
4587
4588static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
4589 const char *line_prefix)
4590{
4591 if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
4592 fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
4593 p->two->mode, show_name ? ' ' : '\n');
4594 if (show_name) {
4595 write_name_quoted(p->two->path, file, '\n');
4596 }
4597 }
4598}
4599
4600static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
4601 const char *line_prefix)
4602{
4603 char *names = pprint_rename(p->one->path, p->two->path);
4604
4605 fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
4606 free(names);
4607 show_mode_change(file, p, 0, line_prefix);
4608}
4609
4610static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
4611{
4612 FILE *file = opt->file;
4613 const char *line_prefix = diff_line_prefix(opt);
4614
4615 switch(p->status) {
4616 case DIFF_STATUS_DELETED:
4617 fputs(line_prefix, file);
4618 show_file_mode_name(file, "delete", p->one);
4619 break;
4620 case DIFF_STATUS_ADDED:
4621 fputs(line_prefix, file);
4622 show_file_mode_name(file, "create", p->two);
4623 break;
4624 case DIFF_STATUS_COPIED:
4625 fputs(line_prefix, file);
4626 show_rename_copy(file, "copy", p, line_prefix);
4627 break;
4628 case DIFF_STATUS_RENAMED:
4629 fputs(line_prefix, file);
4630 show_rename_copy(file, "rename", p, line_prefix);
4631 break;
4632 default:
4633 if (p->score) {
4634 fprintf(file, "%s rewrite ", line_prefix);
4635 write_name_quoted(p->two->path, file, ' ');
4636 fprintf(file, "(%d%%)\n", similarity_index(p));
4637 }
4638 show_mode_change(file, p, !p->score, line_prefix);
4639 break;
4640 }
4641}
4642
4643struct patch_id_t {
4644 git_SHA_CTX *ctx;
4645 int patchlen;
4646};
4647
4648static int remove_space(char *line, int len)
4649{
4650 int i;
4651 char *dst = line;
4652 unsigned char c;
4653
4654 for (i = 0; i < len; i++)
4655 if (!isspace((c = line[i])))
4656 *dst++ = c;
4657
4658 return dst - line;
4659}
4660
4661static void patch_id_consume(void *priv, char *line, unsigned long len)
4662{
4663 struct patch_id_t *data = priv;
4664 int new_len;
4665
4666 /* Ignore line numbers when computing the SHA1 of the patch */
4667 if (starts_with(line, "@@ -"))
4668 return;
4669
4670 new_len = remove_space(line, len);
4671
4672 git_SHA1_Update(data->ctx, line, new_len);
4673 data->patchlen += new_len;
4674}
4675
4676static void patch_id_add_string(git_SHA_CTX *ctx, const char *str)
4677{
4678 git_SHA1_Update(ctx, str, strlen(str));
4679}
4680
4681static void patch_id_add_mode(git_SHA_CTX *ctx, unsigned mode)
4682{
4683 /* large enough for 2^32 in octal */
4684 char buf[12];
4685 int len = xsnprintf(buf, sizeof(buf), "%06o", mode);
4686 git_SHA1_Update(ctx, buf, len);
4687}
4688
4689/* returns 0 upon success, and writes result into sha1 */
4690static int diff_get_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
4691{
4692 struct diff_queue_struct *q = &diff_queued_diff;
4693 int i;
4694 git_SHA_CTX ctx;
4695 struct patch_id_t data;
4696
4697 git_SHA1_Init(&ctx);
4698 memset(&data, 0, sizeof(struct patch_id_t));
4699 data.ctx = &ctx;
4700
4701 for (i = 0; i < q->nr; i++) {
4702 xpparam_t xpp;
4703 xdemitconf_t xecfg;
4704 mmfile_t mf1, mf2;
4705 struct diff_filepair *p = q->queue[i];
4706 int len1, len2;
4707
4708 memset(&xpp, 0, sizeof(xpp));
4709 memset(&xecfg, 0, sizeof(xecfg));
4710 if (p->status == 0)
4711 return error("internal diff status error");
4712 if (p->status == DIFF_STATUS_UNKNOWN)
4713 continue;
4714 if (diff_unmodified_pair(p))
4715 continue;
4716 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4717 (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4718 continue;
4719 if (DIFF_PAIR_UNMERGED(p))
4720 continue;
4721
4722 diff_fill_oid_info(p->one);
4723 diff_fill_oid_info(p->two);
4724
4725 len1 = remove_space(p->one->path, strlen(p->one->path));
4726 len2 = remove_space(p->two->path, strlen(p->two->path));
4727 patch_id_add_string(&ctx, "diff--git");
4728 patch_id_add_string(&ctx, "a/");
4729 git_SHA1_Update(&ctx, p->one->path, len1);
4730 patch_id_add_string(&ctx, "b/");
4731 git_SHA1_Update(&ctx, p->two->path, len2);
4732
4733 if (p->one->mode == 0) {
4734 patch_id_add_string(&ctx, "newfilemode");
4735 patch_id_add_mode(&ctx, p->two->mode);
4736 patch_id_add_string(&ctx, "---/dev/null");
4737 patch_id_add_string(&ctx, "+++b/");
4738 git_SHA1_Update(&ctx, p->two->path, len2);
4739 } else if (p->two->mode == 0) {
4740 patch_id_add_string(&ctx, "deletedfilemode");
4741 patch_id_add_mode(&ctx, p->one->mode);
4742 patch_id_add_string(&ctx, "---a/");
4743 git_SHA1_Update(&ctx, p->one->path, len1);
4744 patch_id_add_string(&ctx, "+++/dev/null");
4745 } else {
4746 patch_id_add_string(&ctx, "---a/");
4747 git_SHA1_Update(&ctx, p->one->path, len1);
4748 patch_id_add_string(&ctx, "+++b/");
4749 git_SHA1_Update(&ctx, p->two->path, len2);
4750 }
4751
4752 if (diff_header_only)
4753 continue;
4754
4755 if (fill_mmfile(&mf1, p->one) < 0 ||
4756 fill_mmfile(&mf2, p->two) < 0)
4757 return error("unable to read files to diff");
4758
4759 if (diff_filespec_is_binary(p->one) ||
4760 diff_filespec_is_binary(p->two)) {
4761 git_SHA1_Update(&ctx, oid_to_hex(&p->one->oid),
4762 GIT_SHA1_HEXSZ);
4763 git_SHA1_Update(&ctx, oid_to_hex(&p->two->oid),
4764 GIT_SHA1_HEXSZ);
4765 continue;
4766 }
4767
4768 xpp.flags = 0;
4769 xecfg.ctxlen = 3;
4770 xecfg.flags = 0;
4771 if (xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4772 &xpp, &xecfg))
4773 return error("unable to generate patch-id diff for %s",
4774 p->one->path);
4775 }
4776
4777 git_SHA1_Final(oid->hash, &ctx);
4778 return 0;
4779}
4780
4781int diff_flush_patch_id(struct diff_options *options, struct object_id *oid, int diff_header_only)
4782{
4783 struct diff_queue_struct *q = &diff_queued_diff;
4784 int i;
4785 int result = diff_get_patch_id(options, oid, diff_header_only);
4786
4787 for (i = 0; i < q->nr; i++)
4788 diff_free_filepair(q->queue[i]);
4789
4790 free(q->queue);
4791 DIFF_QUEUE_CLEAR(q);
4792
4793 return result;
4794}
4795
4796static int is_summary_empty(const struct diff_queue_struct *q)
4797{
4798 int i;
4799
4800 for (i = 0; i < q->nr; i++) {
4801 const struct diff_filepair *p = q->queue[i];
4802
4803 switch (p->status) {
4804 case DIFF_STATUS_DELETED:
4805 case DIFF_STATUS_ADDED:
4806 case DIFF_STATUS_COPIED:
4807 case DIFF_STATUS_RENAMED:
4808 return 0;
4809 default:
4810 if (p->score)
4811 return 0;
4812 if (p->one->mode && p->two->mode &&
4813 p->one->mode != p->two->mode)
4814 return 0;
4815 break;
4816 }
4817 }
4818 return 1;
4819}
4820
4821static const char rename_limit_warning[] =
4822N_("inexact rename detection was skipped due to too many files.");
4823
4824static const char degrade_cc_to_c_warning[] =
4825N_("only found copies from modified paths due to too many files.");
4826
4827static const char rename_limit_advice[] =
4828N_("you may want to set your %s variable to at least "
4829 "%d and retry the command.");
4830
4831void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4832{
4833 if (degraded_cc)
4834 warning(_(degrade_cc_to_c_warning));
4835 else if (needed)
4836 warning(_(rename_limit_warning));
4837 else
4838 return;
4839 if (0 < needed && needed < 32767)
4840 warning(_(rename_limit_advice), varname, needed);
4841}
4842
4843static void diff_flush_patch_all_file_pairs(struct diff_options *o)
4844{
4845 int i;
4846 struct diff_queue_struct *q = &diff_queued_diff;
4847
4848 if (WSEH_NEW & WS_RULE_MASK)
4849 die("BUG: WS rules bit mask overlaps with diff symbol flags");
4850
4851 for (i = 0; i < q->nr; i++) {
4852 struct diff_filepair *p = q->queue[i];
4853 if (check_pair_status(p))
4854 diff_flush_patch(p, o);
4855 }
4856}
4857
4858void diff_flush(struct diff_options *options)
4859{
4860 struct diff_queue_struct *q = &diff_queued_diff;
4861 int i, output_format = options->output_format;
4862 int separator = 0;
4863 int dirstat_by_line = 0;
4864
4865 /*
4866 * Order: raw, stat, summary, patch
4867 * or: name/name-status/checkdiff (other bits clear)
4868 */
4869 if (!q->nr)
4870 goto free_queue;
4871
4872 if (output_format & (DIFF_FORMAT_RAW |
4873 DIFF_FORMAT_NAME |
4874 DIFF_FORMAT_NAME_STATUS |
4875 DIFF_FORMAT_CHECKDIFF)) {
4876 for (i = 0; i < q->nr; i++) {
4877 struct diff_filepair *p = q->queue[i];
4878 if (check_pair_status(p))
4879 flush_one_pair(p, options);
4880 }
4881 separator++;
4882 }
4883
4884 if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4885 dirstat_by_line = 1;
4886
4887 if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4888 dirstat_by_line) {
4889 struct diffstat_t diffstat;
4890
4891 memset(&diffstat, 0, sizeof(struct diffstat_t));
4892 for (i = 0; i < q->nr; i++) {
4893 struct diff_filepair *p = q->queue[i];
4894 if (check_pair_status(p))
4895 diff_flush_stat(p, options, &diffstat);
4896 }
4897 if (output_format & DIFF_FORMAT_NUMSTAT)
4898 show_numstat(&diffstat, options);
4899 if (output_format & DIFF_FORMAT_DIFFSTAT)
4900 show_stats(&diffstat, options);
4901 if (output_format & DIFF_FORMAT_SHORTSTAT)
4902 show_shortstats(&diffstat, options);
4903 if (output_format & DIFF_FORMAT_DIRSTAT && dirstat_by_line)
4904 show_dirstat_by_line(&diffstat, options);
4905 free_diffstat_info(&diffstat);
4906 separator++;
4907 }
4908 if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4909 show_dirstat(options);
4910
4911 if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4912 for (i = 0; i < q->nr; i++) {
4913 diff_summary(options, q->queue[i]);
4914 }
4915 separator++;
4916 }
4917
4918 if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4919 DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4920 DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4921 /*
4922 * run diff_flush_patch for the exit status. setting
4923 * options->file to /dev/null should be safe, because we
4924 * aren't supposed to produce any output anyway.
4925 */
4926 if (options->close_file)
4927 fclose(options->file);
4928 options->file = xfopen("/dev/null", "w");
4929 options->close_file = 1;
4930 for (i = 0; i < q->nr; i++) {
4931 struct diff_filepair *p = q->queue[i];
4932 if (check_pair_status(p))
4933 diff_flush_patch(p, options);
4934 if (options->found_changes)
4935 break;
4936 }
4937 }
4938
4939 if (output_format & DIFF_FORMAT_PATCH) {
4940 if (separator) {
4941 emit_diff_symbol(options, DIFF_SYMBOL_SEPARATOR, NULL, 0, 0);
4942 if (options->stat_sep) {
4943 /* attach patch instead of inline */
4944 fputs(options->stat_sep, options->file);
4945 }
4946 }
4947
4948 diff_flush_patch_all_file_pairs(options);
4949 }
4950
4951 if (output_format & DIFF_FORMAT_CALLBACK)
4952 options->format_callback(q, options, options->format_callback_data);
4953
4954 for (i = 0; i < q->nr; i++)
4955 diff_free_filepair(q->queue[i]);
4956free_queue:
4957 free(q->queue);
4958 DIFF_QUEUE_CLEAR(q);
4959 if (options->close_file)
4960 fclose(options->file);
4961
4962 /*
4963 * Report the content-level differences with HAS_CHANGES;
4964 * diff_addremove/diff_change does not set the bit when
4965 * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4966 */
4967 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4968 if (options->found_changes)
4969 DIFF_OPT_SET(options, HAS_CHANGES);
4970 else
4971 DIFF_OPT_CLR(options, HAS_CHANGES);
4972 }
4973}
4974
4975static int match_filter(const struct diff_options *options, const struct diff_filepair *p)
4976{
4977 return (((p->status == DIFF_STATUS_MODIFIED) &&
4978 ((p->score &&
4979 filter_bit_tst(DIFF_STATUS_FILTER_BROKEN, options)) ||
4980 (!p->score &&
4981 filter_bit_tst(DIFF_STATUS_MODIFIED, options)))) ||
4982 ((p->status != DIFF_STATUS_MODIFIED) &&
4983 filter_bit_tst(p->status, options)));
4984}
4985
4986static void diffcore_apply_filter(struct diff_options *options)
4987{
4988 int i;
4989 struct diff_queue_struct *q = &diff_queued_diff;
4990 struct diff_queue_struct outq;
4991
4992 DIFF_QUEUE_CLEAR(&outq);
4993
4994 if (!options->filter)
4995 return;
4996
4997 if (filter_bit_tst(DIFF_STATUS_FILTER_AON, options)) {
4998 int found;
4999 for (i = found = 0; !found && i < q->nr; i++) {
5000 if (match_filter(options, q->queue[i]))
5001 found++;
5002 }
5003 if (found)
5004 return;
5005
5006 /* otherwise we will clear the whole queue
5007 * by copying the empty outq at the end of this
5008 * function, but first clear the current entries
5009 * in the queue.
5010 */
5011 for (i = 0; i < q->nr; i++)
5012 diff_free_filepair(q->queue[i]);
5013 }
5014 else {
5015 /* Only the matching ones */
5016 for (i = 0; i < q->nr; i++) {
5017 struct diff_filepair *p = q->queue[i];
5018 if (match_filter(options, p))
5019 diff_q(&outq, p);
5020 else
5021 diff_free_filepair(p);
5022 }
5023 }
5024 free(q->queue);
5025 *q = outq;
5026}
5027
5028/* Check whether two filespecs with the same mode and size are identical */
5029static int diff_filespec_is_identical(struct diff_filespec *one,
5030 struct diff_filespec *two)
5031{
5032 if (S_ISGITLINK(one->mode))
5033 return 0;
5034 if (diff_populate_filespec(one, 0))
5035 return 0;
5036 if (diff_populate_filespec(two, 0))
5037 return 0;
5038 return !memcmp(one->data, two->data, one->size);
5039}
5040
5041static int diff_filespec_check_stat_unmatch(struct diff_filepair *p)
5042{
5043 if (p->done_skip_stat_unmatch)
5044 return p->skip_stat_unmatch_result;
5045
5046 p->done_skip_stat_unmatch = 1;
5047 p->skip_stat_unmatch_result = 0;
5048 /*
5049 * 1. Entries that come from stat info dirtiness
5050 * always have both sides (iow, not create/delete),
5051 * one side of the object name is unknown, with
5052 * the same mode and size. Keep the ones that
5053 * do not match these criteria. They have real
5054 * differences.
5055 *
5056 * 2. At this point, the file is known to be modified,
5057 * with the same mode and size, and the object
5058 * name of one side is unknown. Need to inspect
5059 * the identical contents.
5060 */
5061 if (!DIFF_FILE_VALID(p->one) || /* (1) */
5062 !DIFF_FILE_VALID(p->two) ||
5063 (p->one->oid_valid && p->two->oid_valid) ||
5064 (p->one->mode != p->two->mode) ||
5065 diff_populate_filespec(p->one, CHECK_SIZE_ONLY) ||
5066 diff_populate_filespec(p->two, CHECK_SIZE_ONLY) ||
5067 (p->one->size != p->two->size) ||
5068 !diff_filespec_is_identical(p->one, p->two)) /* (2) */
5069 p->skip_stat_unmatch_result = 1;
5070 return p->skip_stat_unmatch_result;
5071}
5072
5073static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
5074{
5075 int i;
5076 struct diff_queue_struct *q = &diff_queued_diff;
5077 struct diff_queue_struct outq;
5078 DIFF_QUEUE_CLEAR(&outq);
5079
5080 for (i = 0; i < q->nr; i++) {
5081 struct diff_filepair *p = q->queue[i];
5082
5083 if (diff_filespec_check_stat_unmatch(p))
5084 diff_q(&outq, p);
5085 else {
5086 /*
5087 * The caller can subtract 1 from skip_stat_unmatch
5088 * to determine how many paths were dirty only
5089 * due to stat info mismatch.
5090 */
5091 if (!DIFF_OPT_TST(diffopt, NO_INDEX))
5092 diffopt->skip_stat_unmatch++;
5093 diff_free_filepair(p);
5094 }
5095 }
5096 free(q->queue);
5097 *q = outq;
5098}
5099
5100static int diffnamecmp(const void *a_, const void *b_)
5101{
5102 const struct diff_filepair *a = *((const struct diff_filepair **)a_);
5103 const struct diff_filepair *b = *((const struct diff_filepair **)b_);
5104 const char *name_a, *name_b;
5105
5106 name_a = a->one ? a->one->path : a->two->path;
5107 name_b = b->one ? b->one->path : b->two->path;
5108 return strcmp(name_a, name_b);
5109}
5110
5111void diffcore_fix_diff_index(struct diff_options *options)
5112{
5113 struct diff_queue_struct *q = &diff_queued_diff;
5114 QSORT(q->queue, q->nr, diffnamecmp);
5115}
5116
5117void diffcore_std(struct diff_options *options)
5118{
5119 /* NOTE please keep the following in sync with diff_tree_combined() */
5120 if (options->skip_stat_unmatch)
5121 diffcore_skip_stat_unmatch(options);
5122 if (!options->found_follow) {
5123 /* See try_to_follow_renames() in tree-diff.c */
5124 if (options->break_opt != -1)
5125 diffcore_break(options->break_opt);
5126 if (options->detect_rename)
5127 diffcore_rename(options);
5128 if (options->break_opt != -1)
5129 diffcore_merge_broken();
5130 }
5131 if (options->pickaxe)
5132 diffcore_pickaxe(options);
5133 if (options->orderfile)
5134 diffcore_order(options->orderfile);
5135 if (!options->found_follow)
5136 /* See try_to_follow_renames() in tree-diff.c */
5137 diff_resolve_rename_copy();
5138 diffcore_apply_filter(options);
5139
5140 if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5141 DIFF_OPT_SET(options, HAS_CHANGES);
5142 else
5143 DIFF_OPT_CLR(options, HAS_CHANGES);
5144
5145 options->found_follow = 0;
5146}
5147
5148int diff_result_code(struct diff_options *opt, int status)
5149{
5150 int result = 0;
5151
5152 diff_warn_rename_limit("diff.renameLimit",
5153 opt->needed_rename_limit,
5154 opt->degraded_cc_to_c);
5155 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5156 !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
5157 return status;
5158 if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5159 DIFF_OPT_TST(opt, HAS_CHANGES))
5160 result |= 01;
5161 if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
5162 DIFF_OPT_TST(opt, CHECK_FAILED))
5163 result |= 02;
5164 return result;
5165}
5166
5167int diff_can_quit_early(struct diff_options *opt)
5168{
5169 return (DIFF_OPT_TST(opt, QUICK) &&
5170 !opt->filter &&
5171 DIFF_OPT_TST(opt, HAS_CHANGES));
5172}
5173
5174/*
5175 * Shall changes to this submodule be ignored?
5176 *
5177 * Submodule changes can be configured to be ignored separately for each path,
5178 * but that configuration can be overridden from the command line.
5179 */
5180static int is_submodule_ignored(const char *path, struct diff_options *options)
5181{
5182 int ignored = 0;
5183 unsigned orig_flags = options->flags;
5184 if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
5185 set_diffopt_flags_from_submodule_config(options, path);
5186 if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
5187 ignored = 1;
5188 options->flags = orig_flags;
5189 return ignored;
5190}
5191
5192void diff_addremove(struct diff_options *options,
5193 int addremove, unsigned mode,
5194 const struct object_id *oid,
5195 int oid_valid,
5196 const char *concatpath, unsigned dirty_submodule)
5197{
5198 struct diff_filespec *one, *two;
5199
5200 if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
5201 return;
5202
5203 /* This may look odd, but it is a preparation for
5204 * feeding "there are unchanged files which should
5205 * not produce diffs, but when you are doing copy
5206 * detection you would need them, so here they are"
5207 * entries to the diff-core. They will be prefixed
5208 * with something like '=' or '*' (I haven't decided
5209 * which but should not make any difference).
5210 * Feeding the same new and old to diff_change()
5211 * also has the same effect.
5212 * Before the final output happens, they are pruned after
5213 * merged into rename/copy pairs as appropriate.
5214 */
5215 if (DIFF_OPT_TST(options, REVERSE_DIFF))
5216 addremove = (addremove == '+' ? '-' :
5217 addremove == '-' ? '+' : addremove);
5218
5219 if (options->prefix &&
5220 strncmp(concatpath, options->prefix, options->prefix_length))
5221 return;
5222
5223 one = alloc_filespec(concatpath);
5224 two = alloc_filespec(concatpath);
5225
5226 if (addremove != '+')
5227 fill_filespec(one, oid, oid_valid, mode);
5228 if (addremove != '-') {
5229 fill_filespec(two, oid, oid_valid, mode);
5230 two->dirty_submodule = dirty_submodule;
5231 }
5232
5233 diff_queue(&diff_queued_diff, one, two);
5234 if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5235 DIFF_OPT_SET(options, HAS_CHANGES);
5236}
5237
5238void diff_change(struct diff_options *options,
5239 unsigned old_mode, unsigned new_mode,
5240 const struct object_id *old_oid,
5241 const struct object_id *new_oid,
5242 int old_oid_valid, int new_oid_valid,
5243 const char *concatpath,
5244 unsigned old_dirty_submodule, unsigned new_dirty_submodule)
5245{
5246 struct diff_filespec *one, *two;
5247 struct diff_filepair *p;
5248
5249 if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
5250 is_submodule_ignored(concatpath, options))
5251 return;
5252
5253 if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
5254 SWAP(old_mode, new_mode);
5255 SWAP(old_oid, new_oid);
5256 SWAP(old_oid_valid, new_oid_valid);
5257 SWAP(old_dirty_submodule, new_dirty_submodule);
5258 }
5259
5260 if (options->prefix &&
5261 strncmp(concatpath, options->prefix, options->prefix_length))
5262 return;
5263
5264 one = alloc_filespec(concatpath);
5265 two = alloc_filespec(concatpath);
5266 fill_filespec(one, old_oid, old_oid_valid, old_mode);
5267 fill_filespec(two, new_oid, new_oid_valid, new_mode);
5268 one->dirty_submodule = old_dirty_submodule;
5269 two->dirty_submodule = new_dirty_submodule;
5270 p = diff_queue(&diff_queued_diff, one, two);
5271
5272 if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
5273 return;
5274
5275 if (DIFF_OPT_TST(options, QUICK) && options->skip_stat_unmatch &&
5276 !diff_filespec_check_stat_unmatch(p))
5277 return;
5278
5279 DIFF_OPT_SET(options, HAS_CHANGES);
5280}
5281
5282struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
5283{
5284 struct diff_filepair *pair;
5285 struct diff_filespec *one, *two;
5286
5287 if (options->prefix &&
5288 strncmp(path, options->prefix, options->prefix_length))
5289 return NULL;
5290
5291 one = alloc_filespec(path);
5292 two = alloc_filespec(path);
5293 pair = diff_queue(&diff_queued_diff, one, two);
5294 pair->is_unmerged = 1;
5295 return pair;
5296}
5297
5298static char *run_textconv(const char *pgm, struct diff_filespec *spec,
5299 size_t *outsize)
5300{
5301 struct diff_tempfile *temp;
5302 const char *argv[3];
5303 const char **arg = argv;
5304 struct child_process child = CHILD_PROCESS_INIT;
5305 struct strbuf buf = STRBUF_INIT;
5306 int err = 0;
5307
5308 temp = prepare_temp_file(spec->path, spec);
5309 *arg++ = pgm;
5310 *arg++ = temp->name;
5311 *arg = NULL;
5312
5313 child.use_shell = 1;
5314 child.argv = argv;
5315 child.out = -1;
5316 if (start_command(&child)) {
5317 remove_tempfile();
5318 return NULL;
5319 }
5320
5321 if (strbuf_read(&buf, child.out, 0) < 0)
5322 err = error("error reading from textconv command '%s'", pgm);
5323 close(child.out);
5324
5325 if (finish_command(&child) || err) {
5326 strbuf_release(&buf);
5327 remove_tempfile();
5328 return NULL;
5329 }
5330 remove_tempfile();
5331
5332 return strbuf_detach(&buf, outsize);
5333}
5334
5335size_t fill_textconv(struct userdiff_driver *driver,
5336 struct diff_filespec *df,
5337 char **outbuf)
5338{
5339 size_t size;
5340
5341 if (!driver) {
5342 if (!DIFF_FILE_VALID(df)) {
5343 *outbuf = "";
5344 return 0;
5345 }
5346 if (diff_populate_filespec(df, 0))
5347 die("unable to read files to diff");
5348 *outbuf = df->data;
5349 return df->size;
5350 }
5351
5352 if (!driver->textconv)
5353 die("BUG: fill_textconv called with non-textconv driver");
5354
5355 if (driver->textconv_cache && df->oid_valid) {
5356 *outbuf = notes_cache_get(driver->textconv_cache,
5357 &df->oid,
5358 &size);
5359 if (*outbuf)
5360 return size;
5361 }
5362
5363 *outbuf = run_textconv(driver->textconv, df, &size);
5364 if (!*outbuf)
5365 die("unable to read files to diff");
5366
5367 if (driver->textconv_cache && df->oid_valid) {
5368 /* ignore errors, as we might be in a readonly repository */
5369 notes_cache_put(driver->textconv_cache, &df->oid, *outbuf,
5370 size);
5371 /*
5372 * we could save up changes and flush them all at the end,
5373 * but we would need an extra call after all diffing is done.
5374 * Since generating a cache entry is the slow path anyway,
5375 * this extra overhead probably isn't a big deal.
5376 */
5377 notes_cache_write(driver->textconv_cache);
5378 }
5379
5380 return size;
5381}
5382
5383int textconv_object(const char *path,
5384 unsigned mode,
5385 const struct object_id *oid,
5386 int oid_valid,
5387 char **buf,
5388 unsigned long *buf_size)
5389{
5390 struct diff_filespec *df;
5391 struct userdiff_driver *textconv;
5392
5393 df = alloc_filespec(path);
5394 fill_filespec(df, oid, oid_valid, mode);
5395 textconv = get_textconv(df);
5396 if (!textconv) {
5397 free_filespec(df);
5398 return 0;
5399 }
5400
5401 *buf_size = fill_textconv(textconv, df, buf);
5402 free_filespec(df);
5403 return 1;
5404}
5405
5406void setup_diff_pager(struct diff_options *opt)
5407{
5408 /*
5409 * If the user asked for our exit code, then either they want --quiet
5410 * or --exit-code. We should definitely not bother with a pager in the
5411 * former case, as we will generate no output. Since we still properly
5412 * report our exit code even when a pager is run, we _could_ run a
5413 * pager with --exit-code. But since we have not done so historically,
5414 * and because it is easy to find people oneline advising "git diff
5415 * --exit-code" in hooks and other scripts, we do not do so.
5416 */
5417 if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
5418 check_pager_config("diff") != 0)
5419 setup_pager();
5420}