diff.c: add dimming to moved line detection
[git/git.git] / diff.c
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 "hashmap.h"
20 #include "ll-merge.h"
21 #include "string-list.h"
22 #include "argv-array.h"
23 #include "graph.h"
24
25 #ifdef NO_FAST_WORKING_DIRECTORY
26 #define FAST_WORKING_DIRECTORY 0
27 #else
28 #define FAST_WORKING_DIRECTORY 1
29 #endif
30
31 static int diff_detect_rename_default;
32 static int diff_indent_heuristic = 1;
33 static int diff_rename_limit_default = 400;
34 static int diff_suppress_blank_empty;
35 static int diff_use_color_default = -1;
36 static int diff_color_moved_default;
37 static int diff_context_default = 3;
38 static int diff_interhunk_context_default;
39 static const char *diff_word_regex_cfg;
40 static const char *external_diff_cmd_cfg;
41 static const char *diff_order_file_cfg;
42 int diff_auto_refresh_index = 1;
43 static int diff_mnemonic_prefix;
44 static int diff_no_prefix;
45 static int diff_stat_graph_width;
46 static int diff_dirstat_permille_default = 30;
47 static struct diff_options default_diff_options;
48 static long diff_algorithm;
49 static unsigned ws_error_highlight_default = WSEH_NEW;
50
51 static char diff_colors[][COLOR_MAXLEN] = {
52 GIT_COLOR_RESET,
53 GIT_COLOR_NORMAL, /* CONTEXT */
54 GIT_COLOR_BOLD, /* METAINFO */
55 GIT_COLOR_CYAN, /* FRAGINFO */
56 GIT_COLOR_RED, /* OLD */
57 GIT_COLOR_GREEN, /* NEW */
58 GIT_COLOR_YELLOW, /* COMMIT */
59 GIT_COLOR_BG_RED, /* WHITESPACE */
60 GIT_COLOR_NORMAL, /* FUNCINFO */
61 GIT_COLOR_BOLD_MAGENTA, /* OLD_MOVED */
62 GIT_COLOR_BOLD_BLUE, /* OLD_MOVED ALTERNATIVE */
63 GIT_COLOR_FAINT, /* OLD_MOVED_DIM */
64 GIT_COLOR_FAINT_ITALIC, /* OLD_MOVED_ALTERNATIVE_DIM */
65 GIT_COLOR_BOLD_CYAN, /* NEW_MOVED */
66 GIT_COLOR_BOLD_YELLOW, /* NEW_MOVED ALTERNATIVE */
67 GIT_COLOR_FAINT, /* NEW_MOVED_DIM */
68 GIT_COLOR_FAINT_ITALIC, /* NEW_MOVED_ALTERNATIVE_DIM */
69 };
70
71 static NORETURN void die_want_option(const char *option_name)
72 {
73 die(_("option '%s' requires a value"), option_name);
74 }
75
76 static int parse_diff_color_slot(const char *var)
77 {
78 if (!strcasecmp(var, "context") || !strcasecmp(var, "plain"))
79 return DIFF_CONTEXT;
80 if (!strcasecmp(var, "meta"))
81 return DIFF_METAINFO;
82 if (!strcasecmp(var, "frag"))
83 return DIFF_FRAGINFO;
84 if (!strcasecmp(var, "old"))
85 return DIFF_FILE_OLD;
86 if (!strcasecmp(var, "new"))
87 return DIFF_FILE_NEW;
88 if (!strcasecmp(var, "commit"))
89 return DIFF_COMMIT;
90 if (!strcasecmp(var, "whitespace"))
91 return DIFF_WHITESPACE;
92 if (!strcasecmp(var, "func"))
93 return DIFF_FUNCINFO;
94 if (!strcasecmp(var, "oldmoved"))
95 return DIFF_FILE_OLD_MOVED;
96 if (!strcasecmp(var, "oldmovedalternative"))
97 return DIFF_FILE_OLD_MOVED_ALT;
98 if (!strcasecmp(var, "oldmoveddimmed"))
99 return DIFF_FILE_OLD_MOVED_DIM;
100 if (!strcasecmp(var, "oldmovedalternativedimmed"))
101 return DIFF_FILE_OLD_MOVED_ALT_DIM;
102 if (!strcasecmp(var, "newmoved"))
103 return DIFF_FILE_NEW_MOVED;
104 if (!strcasecmp(var, "newmovedalternative"))
105 return DIFF_FILE_NEW_MOVED_ALT;
106 if (!strcasecmp(var, "newmoveddimmed"))
107 return DIFF_FILE_NEW_MOVED_DIM;
108 if (!strcasecmp(var, "newmovedalternativedimmed"))
109 return DIFF_FILE_NEW_MOVED_ALT_DIM;
110 return -1;
111 }
112
113 static int parse_dirstat_params(struct diff_options *options, const char *params_string,
114 struct strbuf *errmsg)
115 {
116 char *params_copy = xstrdup(params_string);
117 struct string_list params = STRING_LIST_INIT_NODUP;
118 int ret = 0;
119 int i;
120
121 if (*params_copy)
122 string_list_split_in_place(&params, params_copy, ',', -1);
123 for (i = 0; i < params.nr; i++) {
124 const char *p = params.items[i].string;
125 if (!strcmp(p, "changes")) {
126 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
127 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
128 } else if (!strcmp(p, "lines")) {
129 DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
130 DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
131 } else if (!strcmp(p, "files")) {
132 DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
133 DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
134 } else if (!strcmp(p, "noncumulative")) {
135 DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
136 } else if (!strcmp(p, "cumulative")) {
137 DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
138 } else if (isdigit(*p)) {
139 char *end;
140 int permille = strtoul(p, &end, 10) * 10;
141 if (*end == '.' && isdigit(*++end)) {
142 /* only use first digit */
143 permille += *end - '0';
144 /* .. and ignore any further digits */
145 while (isdigit(*++end))
146 ; /* nothing */
147 }
148 if (!*end)
149 options->dirstat_permille = permille;
150 else {
151 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%s'\n"),
152 p);
153 ret++;
154 }
155 } else {
156 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%s'\n"), p);
157 ret++;
158 }
159
160 }
161 string_list_clear(&params, 0);
162 free(params_copy);
163 return ret;
164 }
165
166 static int parse_submodule_params(struct diff_options *options, const char *value)
167 {
168 if (!strcmp(value, "log"))
169 options->submodule_format = DIFF_SUBMODULE_LOG;
170 else if (!strcmp(value, "short"))
171 options->submodule_format = DIFF_SUBMODULE_SHORT;
172 else if (!strcmp(value, "diff"))
173 options->submodule_format = DIFF_SUBMODULE_INLINE_DIFF;
174 else
175 return -1;
176 return 0;
177 }
178
179 static int git_config_rename(const char *var, const char *value)
180 {
181 if (!value)
182 return DIFF_DETECT_RENAME;
183 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
184 return DIFF_DETECT_COPY;
185 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
186 }
187
188 long parse_algorithm_value(const char *value)
189 {
190 if (!value)
191 return -1;
192 else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
193 return 0;
194 else if (!strcasecmp(value, "minimal"))
195 return XDF_NEED_MINIMAL;
196 else if (!strcasecmp(value, "patience"))
197 return XDF_PATIENCE_DIFF;
198 else if (!strcasecmp(value, "histogram"))
199 return XDF_HISTOGRAM_DIFF;
200 return -1;
201 }
202
203 static int parse_one_token(const char **arg, const char *token)
204 {
205 const char *rest;
206 if (skip_prefix(*arg, token, &rest) && (!*rest || *rest == ',')) {
207 *arg = rest;
208 return 1;
209 }
210 return 0;
211 }
212
213 static int parse_ws_error_highlight(const char *arg)
214 {
215 const char *orig_arg = arg;
216 unsigned val = 0;
217
218 while (*arg) {
219 if (parse_one_token(&arg, "none"))
220 val = 0;
221 else if (parse_one_token(&arg, "default"))
222 val = WSEH_NEW;
223 else if (parse_one_token(&arg, "all"))
224 val = WSEH_NEW | WSEH_OLD | WSEH_CONTEXT;
225 else if (parse_one_token(&arg, "new"))
226 val |= WSEH_NEW;
227 else if (parse_one_token(&arg, "old"))
228 val |= WSEH_OLD;
229 else if (parse_one_token(&arg, "context"))
230 val |= WSEH_CONTEXT;
231 else {
232 return -1 - (int)(arg - orig_arg);
233 }
234 if (*arg)
235 arg++;
236 }
237 return val;
238 }
239
240 /*
241 * These are to give UI layer defaults.
242 * The core-level commands such as git-diff-files should
243 * never be affected by the setting of diff.renames
244 * the user happens to have in the configuration file.
245 */
246 void init_diff_ui_defaults(void)
247 {
248 diff_detect_rename_default = 1;
249 }
250
251 int git_diff_heuristic_config(const char *var, const char *value, void *cb)
252 {
253 if (!strcmp(var, "diff.indentheuristic"))
254 diff_indent_heuristic = git_config_bool(var, value);
255 return 0;
256 }
257
258 static int parse_color_moved(const char *arg)
259 {
260 switch (git_parse_maybe_bool(arg)) {
261 case 0:
262 return COLOR_MOVED_NO;
263 case 1:
264 return COLOR_MOVED_DEFAULT;
265 default:
266 break;
267 }
268
269 if (!strcmp(arg, "no"))
270 return COLOR_MOVED_NO;
271 else if (!strcmp(arg, "plain"))
272 return COLOR_MOVED_PLAIN;
273 else if (!strcmp(arg, "zebra"))
274 return COLOR_MOVED_ZEBRA;
275 else if (!strcmp(arg, "default"))
276 return COLOR_MOVED_DEFAULT;
277 else if (!strcmp(arg, "dimmed_zebra"))
278 return COLOR_MOVED_ZEBRA_DIM;
279 else
280 return error(_("color moved setting must be one of 'no', 'default', 'zebra', 'dimmed_zebra', 'plain'"));
281 }
282
283 int git_diff_ui_config(const char *var, const char *value, void *cb)
284 {
285 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
286 diff_use_color_default = git_config_colorbool(var, value);
287 return 0;
288 }
289 if (!strcmp(var, "diff.colormoved")) {
290 int cm = parse_color_moved(value);
291 if (cm < 0)
292 return -1;
293 diff_color_moved_default = cm;
294 return 0;
295 }
296 if (!strcmp(var, "diff.context")) {
297 diff_context_default = git_config_int(var, value);
298 if (diff_context_default < 0)
299 return -1;
300 return 0;
301 }
302 if (!strcmp(var, "diff.interhunkcontext")) {
303 diff_interhunk_context_default = git_config_int(var, value);
304 if (diff_interhunk_context_default < 0)
305 return -1;
306 return 0;
307 }
308 if (!strcmp(var, "diff.renames")) {
309 diff_detect_rename_default = git_config_rename(var, value);
310 return 0;
311 }
312 if (!strcmp(var, "diff.autorefreshindex")) {
313 diff_auto_refresh_index = git_config_bool(var, value);
314 return 0;
315 }
316 if (!strcmp(var, "diff.mnemonicprefix")) {
317 diff_mnemonic_prefix = git_config_bool(var, value);
318 return 0;
319 }
320 if (!strcmp(var, "diff.noprefix")) {
321 diff_no_prefix = git_config_bool(var, value);
322 return 0;
323 }
324 if (!strcmp(var, "diff.statgraphwidth")) {
325 diff_stat_graph_width = git_config_int(var, value);
326 return 0;
327 }
328 if (!strcmp(var, "diff.external"))
329 return git_config_string(&external_diff_cmd_cfg, var, value);
330 if (!strcmp(var, "diff.wordregex"))
331 return git_config_string(&diff_word_regex_cfg, var, value);
332 if (!strcmp(var, "diff.orderfile"))
333 return git_config_pathname(&diff_order_file_cfg, var, value);
334
335 if (!strcmp(var, "diff.ignoresubmodules"))
336 handle_ignore_submodules_arg(&default_diff_options, value);
337
338 if (!strcmp(var, "diff.submodule")) {
339 if (parse_submodule_params(&default_diff_options, value))
340 warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
341 value);
342 return 0;
343 }
344
345 if (!strcmp(var, "diff.algorithm")) {
346 diff_algorithm = parse_algorithm_value(value);
347 if (diff_algorithm < 0)
348 return -1;
349 return 0;
350 }
351
352 if (!strcmp(var, "diff.wserrorhighlight")) {
353 int val = parse_ws_error_highlight(value);
354 if (val < 0)
355 return -1;
356 ws_error_highlight_default = val;
357 return 0;
358 }
359
360 if (git_color_config(var, value, cb) < 0)
361 return -1;
362
363 return git_diff_basic_config(var, value, cb);
364 }
365
366 int git_diff_basic_config(const char *var, const char *value, void *cb)
367 {
368 const char *name;
369
370 if (!strcmp(var, "diff.renamelimit")) {
371 diff_rename_limit_default = git_config_int(var, value);
372 return 0;
373 }
374
375 if (userdiff_config(var, value) < 0)
376 return -1;
377
378 if (skip_prefix(var, "diff.color.", &name) ||
379 skip_prefix(var, "color.diff.", &name)) {
380 int slot = parse_diff_color_slot(name);
381 if (slot < 0)
382 return 0;
383 if (!value)
384 return config_error_nonbool(var);
385 return color_parse(value, diff_colors[slot]);
386 }
387
388 /* like GNU diff's --suppress-blank-empty option */
389 if (!strcmp(var, "diff.suppressblankempty") ||
390 /* for backwards compatibility */
391 !strcmp(var, "diff.suppress-blank-empty")) {
392 diff_suppress_blank_empty = git_config_bool(var, value);
393 return 0;
394 }
395
396 if (!strcmp(var, "diff.dirstat")) {
397 struct strbuf errmsg = STRBUF_INIT;
398 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
399 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
400 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
401 errmsg.buf);
402 strbuf_release(&errmsg);
403 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
404 return 0;
405 }
406
407 if (starts_with(var, "submodule."))
408 return parse_submodule_config_option(var, value);
409
410 if (git_diff_heuristic_config(var, value, cb) < 0)
411 return -1;
412
413 return git_default_config(var, value, cb);
414 }
415
416 static char *quote_two(const char *one, const char *two)
417 {
418 int need_one = quote_c_style(one, NULL, NULL, 1);
419 int need_two = quote_c_style(two, NULL, NULL, 1);
420 struct strbuf res = STRBUF_INIT;
421
422 if (need_one + need_two) {
423 strbuf_addch(&res, '"');
424 quote_c_style(one, &res, NULL, 1);
425 quote_c_style(two, &res, NULL, 1);
426 strbuf_addch(&res, '"');
427 } else {
428 strbuf_addstr(&res, one);
429 strbuf_addstr(&res, two);
430 }
431 return strbuf_detach(&res, NULL);
432 }
433
434 static const char *external_diff(void)
435 {
436 static const char *external_diff_cmd = NULL;
437 static int done_preparing = 0;
438
439 if (done_preparing)
440 return external_diff_cmd;
441 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
442 if (!external_diff_cmd)
443 external_diff_cmd = external_diff_cmd_cfg;
444 done_preparing = 1;
445 return external_diff_cmd;
446 }
447
448 /*
449 * Keep track of files used for diffing. Sometimes such an entry
450 * refers to a temporary file, sometimes to an existing file, and
451 * sometimes to "/dev/null".
452 */
453 static struct diff_tempfile {
454 /*
455 * filename external diff should read from, or NULL if this
456 * entry is currently not in use:
457 */
458 const char *name;
459
460 char hex[GIT_MAX_HEXSZ + 1];
461 char mode[10];
462
463 /*
464 * If this diff_tempfile instance refers to a temporary file,
465 * this tempfile object is used to manage its lifetime.
466 */
467 struct tempfile tempfile;
468 } diff_temp[2];
469
470 typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
471
472 struct emit_callback {
473 int color_diff;
474 unsigned ws_rule;
475 int blank_at_eof_in_preimage;
476 int blank_at_eof_in_postimage;
477 int lno_in_preimage;
478 int lno_in_postimage;
479 sane_truncate_fn truncate;
480 const char **label_path;
481 struct diff_words_data *diff_words;
482 struct diff_options *opt;
483 struct strbuf *header;
484 };
485
486 static int count_lines(const char *data, int size)
487 {
488 int count, ch, completely_empty = 1, nl_just_seen = 0;
489 count = 0;
490 while (0 < size--) {
491 ch = *data++;
492 if (ch == '\n') {
493 count++;
494 nl_just_seen = 1;
495 completely_empty = 0;
496 }
497 else {
498 nl_just_seen = 0;
499 completely_empty = 0;
500 }
501 }
502 if (completely_empty)
503 return 0;
504 if (!nl_just_seen)
505 count++; /* no trailing newline */
506 return count;
507 }
508
509 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
510 {
511 if (!DIFF_FILE_VALID(one)) {
512 mf->ptr = (char *)""; /* does not matter */
513 mf->size = 0;
514 return 0;
515 }
516 else if (diff_populate_filespec(one, 0))
517 return -1;
518
519 mf->ptr = one->data;
520 mf->size = one->size;
521 return 0;
522 }
523
524 /* like fill_mmfile, but only for size, so we can avoid retrieving blob */
525 static unsigned long diff_filespec_size(struct diff_filespec *one)
526 {
527 if (!DIFF_FILE_VALID(one))
528 return 0;
529 diff_populate_filespec(one, CHECK_SIZE_ONLY);
530 return one->size;
531 }
532
533 static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
534 {
535 char *ptr = mf->ptr;
536 long size = mf->size;
537 int cnt = 0;
538
539 if (!size)
540 return cnt;
541 ptr += size - 1; /* pointing at the very end */
542 if (*ptr != '\n')
543 ; /* incomplete line */
544 else
545 ptr--; /* skip the last LF */
546 while (mf->ptr < ptr) {
547 char *prev_eol;
548 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
549 if (*prev_eol == '\n')
550 break;
551 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
552 break;
553 cnt++;
554 ptr = prev_eol - 1;
555 }
556 return cnt;
557 }
558
559 static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
560 struct emit_callback *ecbdata)
561 {
562 int l1, l2, at;
563 unsigned ws_rule = ecbdata->ws_rule;
564 l1 = count_trailing_blank(mf1, ws_rule);
565 l2 = count_trailing_blank(mf2, ws_rule);
566 if (l2 <= l1) {
567 ecbdata->blank_at_eof_in_preimage = 0;
568 ecbdata->blank_at_eof_in_postimage = 0;
569 return;
570 }
571 at = count_lines(mf1->ptr, mf1->size);
572 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
573
574 at = count_lines(mf2->ptr, mf2->size);
575 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
576 }
577
578 static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
579 int first, const char *line, int len)
580 {
581 int has_trailing_newline, has_trailing_carriage_return;
582 int nofirst;
583 FILE *file = o->file;
584
585 fputs(diff_line_prefix(o), file);
586
587 if (len == 0) {
588 has_trailing_newline = (first == '\n');
589 has_trailing_carriage_return = (!has_trailing_newline &&
590 (first == '\r'));
591 nofirst = has_trailing_newline || has_trailing_carriage_return;
592 } else {
593 has_trailing_newline = (len > 0 && line[len-1] == '\n');
594 if (has_trailing_newline)
595 len--;
596 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
597 if (has_trailing_carriage_return)
598 len--;
599 nofirst = 0;
600 }
601
602 if (len || !nofirst) {
603 fputs(set, file);
604 if (!nofirst)
605 fputc(first, file);
606 fwrite(line, len, 1, file);
607 fputs(reset, file);
608 }
609 if (has_trailing_carriage_return)
610 fputc('\r', file);
611 if (has_trailing_newline)
612 fputc('\n', file);
613 }
614
615 static void emit_line(struct diff_options *o, const char *set, const char *reset,
616 const char *line, int len)
617 {
618 emit_line_0(o, set, reset, line[0], line+1, len-1);
619 }
620
621 enum diff_symbol {
622 DIFF_SYMBOL_BINARY_DIFF_HEADER,
623 DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA,
624 DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL,
625 DIFF_SYMBOL_BINARY_DIFF_BODY,
626 DIFF_SYMBOL_BINARY_DIFF_FOOTER,
627 DIFF_SYMBOL_STATS_SUMMARY_NO_FILES,
628 DIFF_SYMBOL_STATS_SUMMARY_ABBREV,
629 DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES,
630 DIFF_SYMBOL_STATS_LINE,
631 DIFF_SYMBOL_WORD_DIFF,
632 DIFF_SYMBOL_STAT_SEP,
633 DIFF_SYMBOL_SUMMARY,
634 DIFF_SYMBOL_SUBMODULE_ADD,
635 DIFF_SYMBOL_SUBMODULE_DEL,
636 DIFF_SYMBOL_SUBMODULE_UNTRACKED,
637 DIFF_SYMBOL_SUBMODULE_MODIFIED,
638 DIFF_SYMBOL_SUBMODULE_HEADER,
639 DIFF_SYMBOL_SUBMODULE_ERROR,
640 DIFF_SYMBOL_SUBMODULE_PIPETHROUGH,
641 DIFF_SYMBOL_REWRITE_DIFF,
642 DIFF_SYMBOL_BINARY_FILES,
643 DIFF_SYMBOL_HEADER,
644 DIFF_SYMBOL_FILEPAIR_PLUS,
645 DIFF_SYMBOL_FILEPAIR_MINUS,
646 DIFF_SYMBOL_WORDS_PORCELAIN,
647 DIFF_SYMBOL_WORDS,
648 DIFF_SYMBOL_CONTEXT,
649 DIFF_SYMBOL_CONTEXT_INCOMPLETE,
650 DIFF_SYMBOL_PLUS,
651 DIFF_SYMBOL_MINUS,
652 DIFF_SYMBOL_NO_LF_EOF,
653 DIFF_SYMBOL_CONTEXT_FRAGINFO,
654 DIFF_SYMBOL_CONTEXT_MARKER,
655 DIFF_SYMBOL_SEPARATOR
656 };
657 /*
658 * Flags for content lines:
659 * 0..12 are whitespace rules
660 * 13-15 are WSEH_NEW | WSEH_OLD | WSEH_CONTEXT
661 * 16 is marking if the line is blank at EOF
662 */
663 #define DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF (1<<16)
664 #define DIFF_SYMBOL_MOVED_LINE (1<<17)
665 #define DIFF_SYMBOL_MOVED_LINE_ALT (1<<18)
666 #define DIFF_SYMBOL_MOVED_LINE_UNINTERESTING (1<<19)
667 #define DIFF_SYMBOL_CONTENT_WS_MASK (WSEH_NEW | WSEH_OLD | WSEH_CONTEXT | WS_RULE_MASK)
668
669 /*
670 * This struct is used when we need to buffer the output of the diff output.
671 *
672 * NEEDSWORK: Instead of storing a copy of the line, add an offset pointer
673 * into the pre/post image file. This pointer could be a union with the
674 * line pointer. By storing an offset into the file instead of the literal line,
675 * we can decrease the memory footprint for the buffered output. At first we
676 * may want to only have indirection for the content lines, but we could also
677 * enhance the state for emitting prefabricated lines, e.g. the similarity
678 * score line or hunk/file headers would only need to store a number or path
679 * and then the output can be constructed later on depending on state.
680 */
681 struct emitted_diff_symbol {
682 const char *line;
683 int len;
684 int flags;
685 enum diff_symbol s;
686 };
687 #define EMITTED_DIFF_SYMBOL_INIT {NULL}
688
689 struct emitted_diff_symbols {
690 struct emitted_diff_symbol *buf;
691 int nr, alloc;
692 };
693 #define EMITTED_DIFF_SYMBOLS_INIT {NULL, 0, 0}
694
695 static void append_emitted_diff_symbol(struct diff_options *o,
696 struct emitted_diff_symbol *e)
697 {
698 struct emitted_diff_symbol *f;
699
700 ALLOC_GROW(o->emitted_symbols->buf,
701 o->emitted_symbols->nr + 1,
702 o->emitted_symbols->alloc);
703 f = &o->emitted_symbols->buf[o->emitted_symbols->nr++];
704
705 memcpy(f, e, sizeof(struct emitted_diff_symbol));
706 f->line = e->line ? xmemdupz(e->line, e->len) : NULL;
707 }
708
709 struct moved_entry {
710 struct hashmap_entry ent;
711 const struct emitted_diff_symbol *es;
712 struct moved_entry *next_line;
713 };
714
715 static int next_byte(const char **cp, const char **endp,
716 const struct diff_options *diffopt)
717 {
718 int retval;
719
720 if (*cp > *endp)
721 return -1;
722
723 if (DIFF_XDL_TST(diffopt, IGNORE_WHITESPACE_CHANGE)) {
724 while (*cp < *endp && isspace(**cp))
725 (*cp)++;
726 /*
727 * After skipping a couple of whitespaces, we still have to
728 * account for one space.
729 */
730 return (int)' ';
731 }
732
733 if (DIFF_XDL_TST(diffopt, IGNORE_WHITESPACE)) {
734 while (*cp < *endp && isspace(**cp))
735 (*cp)++;
736 /* return the first non-ws character via the usual below */
737 }
738
739 retval = (unsigned char)(**cp);
740 (*cp)++;
741 return retval;
742 }
743
744 static int moved_entry_cmp(const struct diff_options *diffopt,
745 const struct moved_entry *a,
746 const struct moved_entry *b,
747 const void *keydata)
748 {
749 const char *ap = a->es->line, *ae = a->es->line + a->es->len;
750 const char *bp = b->es->line, *be = b->es->line + b->es->len;
751
752 if (!(diffopt->xdl_opts & XDF_WHITESPACE_FLAGS))
753 return a->es->len != b->es->len || memcmp(ap, bp, a->es->len);
754
755 if (DIFF_XDL_TST(diffopt, IGNORE_WHITESPACE_AT_EOL)) {
756 while (ae > ap && isspace(*ae))
757 ae--;
758 while (be > bp && isspace(*be))
759 be--;
760 }
761
762 while (1) {
763 int ca, cb;
764 ca = next_byte(&ap, &ae, diffopt);
765 cb = next_byte(&bp, &be, diffopt);
766 if (ca != cb)
767 return 1;
768 if (ca < 0)
769 return 0;
770 }
771 }
772
773 static unsigned get_string_hash(struct emitted_diff_symbol *es, struct diff_options *o)
774 {
775 if (o->xdl_opts & XDF_WHITESPACE_FLAGS) {
776 static struct strbuf sb = STRBUF_INIT;
777 const char *ap = es->line, *ae = es->line + es->len;
778 int c;
779
780 strbuf_reset(&sb);
781 while (ae > ap && isspace(*ae))
782 ae--;
783 while ((c = next_byte(&ap, &ae, o)) > 0)
784 strbuf_addch(&sb, c);
785
786 return memhash(sb.buf, sb.len);
787 } else {
788 return memhash(es->line, es->len);
789 }
790 }
791
792 static struct moved_entry *prepare_entry(struct diff_options *o,
793 int line_no)
794 {
795 struct moved_entry *ret = xmalloc(sizeof(*ret));
796 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[line_no];
797
798 ret->ent.hash = get_string_hash(l, o);
799 ret->es = l;
800 ret->next_line = NULL;
801
802 return ret;
803 }
804
805 static void add_lines_to_move_detection(struct diff_options *o,
806 struct hashmap *add_lines,
807 struct hashmap *del_lines)
808 {
809 struct moved_entry *prev_line = NULL;
810
811 int n;
812 for (n = 0; n < o->emitted_symbols->nr; n++) {
813 struct hashmap *hm;
814 struct moved_entry *key;
815
816 switch (o->emitted_symbols->buf[n].s) {
817 case DIFF_SYMBOL_PLUS:
818 hm = add_lines;
819 break;
820 case DIFF_SYMBOL_MINUS:
821 hm = del_lines;
822 break;
823 default:
824 prev_line = NULL;
825 continue;
826 }
827
828 key = prepare_entry(o, n);
829 if (prev_line && prev_line->es->s == o->emitted_symbols->buf[n].s)
830 prev_line->next_line = key;
831
832 hashmap_add(hm, key);
833 prev_line = key;
834 }
835 }
836
837 static int shrink_potential_moved_blocks(struct moved_entry **pmb,
838 int pmb_nr)
839 {
840 int lp, rp;
841
842 /* Shrink the set of potential block to the remaining running */
843 for (lp = 0, rp = pmb_nr - 1; lp <= rp;) {
844 while (lp < pmb_nr && pmb[lp])
845 lp++;
846 /* lp points at the first NULL now */
847
848 while (rp > -1 && !pmb[rp])
849 rp--;
850 /* rp points at the last non-NULL */
851
852 if (lp < pmb_nr && rp > -1 && lp < rp) {
853 pmb[lp] = pmb[rp];
854 pmb[rp] = NULL;
855 rp--;
856 lp++;
857 }
858 }
859
860 /* Remember the number of running sets */
861 return rp + 1;
862 }
863
864 /* Find blocks of moved code, delegate actual coloring decision to helper */
865 static void mark_color_as_moved(struct diff_options *o,
866 struct hashmap *add_lines,
867 struct hashmap *del_lines)
868 {
869 struct moved_entry **pmb = NULL; /* potentially moved blocks */
870 int pmb_nr = 0, pmb_alloc = 0;
871 int n, flipped_block = 1, block_length = 0;
872
873
874 for (n = 0; n < o->emitted_symbols->nr; n++) {
875 struct hashmap *hm = NULL;
876 struct moved_entry *key;
877 struct moved_entry *match = NULL;
878 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[n];
879 int i;
880
881 switch (l->s) {
882 case DIFF_SYMBOL_PLUS:
883 hm = del_lines;
884 key = prepare_entry(o, n);
885 match = hashmap_get(hm, key, o);
886 free(key);
887 break;
888 case DIFF_SYMBOL_MINUS:
889 hm = add_lines;
890 key = prepare_entry(o, n);
891 match = hashmap_get(hm, key, o);
892 free(key);
893 break;
894 default:
895 flipped_block = 1;
896 }
897
898 if (!match) {
899 if (block_length < COLOR_MOVED_MIN_BLOCK_LENGTH &&
900 o->color_moved != COLOR_MOVED_PLAIN) {
901 for (i = 0; i < block_length + 1; i++) {
902 l = &o->emitted_symbols->buf[n - i];
903 l->flags &= ~DIFF_SYMBOL_MOVED_LINE;
904 }
905 }
906 pmb_nr = 0;
907 block_length = 0;
908 continue;
909 }
910
911 l->flags |= DIFF_SYMBOL_MOVED_LINE;
912 block_length++;
913
914 if (o->color_moved == COLOR_MOVED_PLAIN)
915 continue;
916
917 /* Check any potential block runs, advance each or nullify */
918 for (i = 0; i < pmb_nr; i++) {
919 struct moved_entry *p = pmb[i];
920 struct moved_entry *pnext = (p && p->next_line) ?
921 p->next_line : NULL;
922 if (pnext && !hm->cmpfn(o, pnext, match, NULL)) {
923 pmb[i] = p->next_line;
924 } else {
925 pmb[i] = NULL;
926 }
927 }
928
929 pmb_nr = shrink_potential_moved_blocks(pmb, pmb_nr);
930
931 if (pmb_nr == 0) {
932 /*
933 * The current line is the start of a new block.
934 * Setup the set of potential blocks.
935 */
936 for (; match; match = hashmap_get_next(hm, match)) {
937 ALLOC_GROW(pmb, pmb_nr + 1, pmb_alloc);
938 pmb[pmb_nr++] = match;
939 }
940
941 flipped_block = (flipped_block + 1) % 2;
942 }
943
944 if (flipped_block)
945 l->flags |= DIFF_SYMBOL_MOVED_LINE_ALT;
946 }
947
948 free(pmb);
949 }
950
951 #define DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK \
952 (DIFF_SYMBOL_MOVED_LINE | DIFF_SYMBOL_MOVED_LINE_ALT)
953 static void dim_moved_lines(struct diff_options *o)
954 {
955 int n;
956 for (n = 0; n < o->emitted_symbols->nr; n++) {
957 struct emitted_diff_symbol *prev = (n != 0) ?
958 &o->emitted_symbols->buf[n - 1] : NULL;
959 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[n];
960 struct emitted_diff_symbol *next =
961 (n < o->emitted_symbols->nr - 1) ?
962 &o->emitted_symbols->buf[n + 1] : NULL;
963
964 /* Not a plus or minus line? */
965 if (l->s != DIFF_SYMBOL_PLUS && l->s != DIFF_SYMBOL_MINUS)
966 continue;
967
968 /* Not a moved line? */
969 if (!(l->flags & DIFF_SYMBOL_MOVED_LINE))
970 continue;
971
972 /*
973 * If prev or next are not a plus or minus line,
974 * pretend they don't exist
975 */
976 if (prev && prev->s != DIFF_SYMBOL_PLUS &&
977 prev->s != DIFF_SYMBOL_MINUS)
978 prev = NULL;
979 if (next && next->s != DIFF_SYMBOL_PLUS &&
980 next->s != DIFF_SYMBOL_MINUS)
981 next = NULL;
982
983 /* Inside a block? */
984 if ((prev &&
985 (prev->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK) ==
986 (l->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK)) &&
987 (next &&
988 (next->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK) ==
989 (l->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK))) {
990 l->flags |= DIFF_SYMBOL_MOVED_LINE_UNINTERESTING;
991 continue;
992 }
993
994 /* Check if we are at an interesting bound: */
995 if (prev && (prev->flags & DIFF_SYMBOL_MOVED_LINE) &&
996 (prev->flags & DIFF_SYMBOL_MOVED_LINE_ALT) !=
997 (l->flags & DIFF_SYMBOL_MOVED_LINE_ALT))
998 continue;
999 if (next && (next->flags & DIFF_SYMBOL_MOVED_LINE) &&
1000 (next->flags & DIFF_SYMBOL_MOVED_LINE_ALT) !=
1001 (l->flags & DIFF_SYMBOL_MOVED_LINE_ALT))
1002 continue;
1003
1004 /*
1005 * The boundary to prev and next are not interesting,
1006 * so this line is not interesting as a whole
1007 */
1008 l->flags |= DIFF_SYMBOL_MOVED_LINE_UNINTERESTING;
1009 }
1010 }
1011
1012 static void emit_line_ws_markup(struct diff_options *o,
1013 const char *set, const char *reset,
1014 const char *line, int len, char sign,
1015 unsigned ws_rule, int blank_at_eof)
1016 {
1017 const char *ws = NULL;
1018
1019 if (o->ws_error_highlight & ws_rule) {
1020 ws = diff_get_color_opt(o, DIFF_WHITESPACE);
1021 if (!*ws)
1022 ws = NULL;
1023 }
1024
1025 if (!ws)
1026 emit_line_0(o, set, reset, sign, line, len);
1027 else if (blank_at_eof)
1028 /* Blank line at EOF - paint '+' as well */
1029 emit_line_0(o, ws, reset, sign, line, len);
1030 else {
1031 /* Emit just the prefix, then the rest. */
1032 emit_line_0(o, set, reset, sign, "", 0);
1033 ws_check_emit(line, len, ws_rule,
1034 o->file, set, reset, ws);
1035 }
1036 }
1037
1038 static void emit_diff_symbol_from_struct(struct diff_options *o,
1039 struct emitted_diff_symbol *eds)
1040 {
1041 static const char *nneof = " No newline at end of file\n";
1042 const char *context, *reset, *set, *meta, *fraginfo;
1043 struct strbuf sb = STRBUF_INIT;
1044
1045 enum diff_symbol s = eds->s;
1046 const char *line = eds->line;
1047 int len = eds->len;
1048 unsigned flags = eds->flags;
1049
1050 switch (s) {
1051 case DIFF_SYMBOL_NO_LF_EOF:
1052 context = diff_get_color_opt(o, DIFF_CONTEXT);
1053 reset = diff_get_color_opt(o, DIFF_RESET);
1054 putc('\n', o->file);
1055 emit_line_0(o, context, reset, '\\',
1056 nneof, strlen(nneof));
1057 break;
1058 case DIFF_SYMBOL_SUBMODULE_HEADER:
1059 case DIFF_SYMBOL_SUBMODULE_ERROR:
1060 case DIFF_SYMBOL_SUBMODULE_PIPETHROUGH:
1061 case DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES:
1062 case DIFF_SYMBOL_SUMMARY:
1063 case DIFF_SYMBOL_STATS_LINE:
1064 case DIFF_SYMBOL_BINARY_DIFF_BODY:
1065 case DIFF_SYMBOL_CONTEXT_FRAGINFO:
1066 emit_line(o, "", "", line, len);
1067 break;
1068 case DIFF_SYMBOL_CONTEXT_INCOMPLETE:
1069 case DIFF_SYMBOL_CONTEXT_MARKER:
1070 context = diff_get_color_opt(o, DIFF_CONTEXT);
1071 reset = diff_get_color_opt(o, DIFF_RESET);
1072 emit_line(o, context, reset, line, len);
1073 break;
1074 case DIFF_SYMBOL_SEPARATOR:
1075 fprintf(o->file, "%s%c",
1076 diff_line_prefix(o),
1077 o->line_termination);
1078 break;
1079 case DIFF_SYMBOL_CONTEXT:
1080 set = diff_get_color_opt(o, DIFF_CONTEXT);
1081 reset = diff_get_color_opt(o, DIFF_RESET);
1082 emit_line_ws_markup(o, set, reset, line, len, ' ',
1083 flags & (DIFF_SYMBOL_CONTENT_WS_MASK), 0);
1084 break;
1085 case DIFF_SYMBOL_PLUS:
1086 switch (flags & (DIFF_SYMBOL_MOVED_LINE |
1087 DIFF_SYMBOL_MOVED_LINE_ALT |
1088 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING)) {
1089 case DIFF_SYMBOL_MOVED_LINE |
1090 DIFF_SYMBOL_MOVED_LINE_ALT |
1091 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1092 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_ALT_DIM);
1093 break;
1094 case DIFF_SYMBOL_MOVED_LINE |
1095 DIFF_SYMBOL_MOVED_LINE_ALT:
1096 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_ALT);
1097 break;
1098 case DIFF_SYMBOL_MOVED_LINE |
1099 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1100 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_DIM);
1101 break;
1102 case DIFF_SYMBOL_MOVED_LINE:
1103 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED);
1104 break;
1105 default:
1106 set = diff_get_color_opt(o, DIFF_FILE_NEW);
1107 }
1108 reset = diff_get_color_opt(o, DIFF_RESET);
1109 emit_line_ws_markup(o, set, reset, line, len, '+',
1110 flags & DIFF_SYMBOL_CONTENT_WS_MASK,
1111 flags & DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF);
1112 break;
1113 case DIFF_SYMBOL_MINUS:
1114 switch (flags & (DIFF_SYMBOL_MOVED_LINE |
1115 DIFF_SYMBOL_MOVED_LINE_ALT |
1116 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING)) {
1117 case DIFF_SYMBOL_MOVED_LINE |
1118 DIFF_SYMBOL_MOVED_LINE_ALT |
1119 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1120 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_ALT_DIM);
1121 break;
1122 case DIFF_SYMBOL_MOVED_LINE |
1123 DIFF_SYMBOL_MOVED_LINE_ALT:
1124 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_ALT);
1125 break;
1126 case DIFF_SYMBOL_MOVED_LINE |
1127 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1128 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_DIM);
1129 break;
1130 case DIFF_SYMBOL_MOVED_LINE:
1131 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED);
1132 break;
1133 default:
1134 set = diff_get_color_opt(o, DIFF_FILE_OLD);
1135 }
1136 reset = diff_get_color_opt(o, DIFF_RESET);
1137 emit_line_ws_markup(o, set, reset, line, len, '-',
1138 flags & DIFF_SYMBOL_CONTENT_WS_MASK, 0);
1139 break;
1140 case DIFF_SYMBOL_WORDS_PORCELAIN:
1141 context = diff_get_color_opt(o, DIFF_CONTEXT);
1142 reset = diff_get_color_opt(o, DIFF_RESET);
1143 emit_line(o, context, reset, line, len);
1144 fputs("~\n", o->file);
1145 break;
1146 case DIFF_SYMBOL_WORDS:
1147 context = diff_get_color_opt(o, DIFF_CONTEXT);
1148 reset = diff_get_color_opt(o, DIFF_RESET);
1149 /*
1150 * Skip the prefix character, if any. With
1151 * diff_suppress_blank_empty, there may be
1152 * none.
1153 */
1154 if (line[0] != '\n') {
1155 line++;
1156 len--;
1157 }
1158 emit_line(o, context, reset, line, len);
1159 break;
1160 case DIFF_SYMBOL_FILEPAIR_PLUS:
1161 meta = diff_get_color_opt(o, DIFF_METAINFO);
1162 reset = diff_get_color_opt(o, DIFF_RESET);
1163 fprintf(o->file, "%s%s+++ %s%s%s\n", diff_line_prefix(o), meta,
1164 line, reset,
1165 strchr(line, ' ') ? "\t" : "");
1166 break;
1167 case DIFF_SYMBOL_FILEPAIR_MINUS:
1168 meta = diff_get_color_opt(o, DIFF_METAINFO);
1169 reset = diff_get_color_opt(o, DIFF_RESET);
1170 fprintf(o->file, "%s%s--- %s%s%s\n", diff_line_prefix(o), meta,
1171 line, reset,
1172 strchr(line, ' ') ? "\t" : "");
1173 break;
1174 case DIFF_SYMBOL_BINARY_FILES:
1175 case DIFF_SYMBOL_HEADER:
1176 fprintf(o->file, "%s", line);
1177 break;
1178 case DIFF_SYMBOL_BINARY_DIFF_HEADER:
1179 fprintf(o->file, "%sGIT binary patch\n", diff_line_prefix(o));
1180 break;
1181 case DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA:
1182 fprintf(o->file, "%sdelta %s\n", diff_line_prefix(o), line);
1183 break;
1184 case DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL:
1185 fprintf(o->file, "%sliteral %s\n", diff_line_prefix(o), line);
1186 break;
1187 case DIFF_SYMBOL_BINARY_DIFF_FOOTER:
1188 fputs(diff_line_prefix(o), o->file);
1189 fputc('\n', o->file);
1190 break;
1191 case DIFF_SYMBOL_REWRITE_DIFF:
1192 fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
1193 reset = diff_get_color_opt(o, DIFF_RESET);
1194 emit_line(o, fraginfo, reset, line, len);
1195 break;
1196 case DIFF_SYMBOL_SUBMODULE_ADD:
1197 set = diff_get_color_opt(o, DIFF_FILE_NEW);
1198 reset = diff_get_color_opt(o, DIFF_RESET);
1199 emit_line(o, set, reset, line, len);
1200 break;
1201 case DIFF_SYMBOL_SUBMODULE_DEL:
1202 set = diff_get_color_opt(o, DIFF_FILE_OLD);
1203 reset = diff_get_color_opt(o, DIFF_RESET);
1204 emit_line(o, set, reset, line, len);
1205 break;
1206 case DIFF_SYMBOL_SUBMODULE_UNTRACKED:
1207 fprintf(o->file, "%sSubmodule %s contains untracked content\n",
1208 diff_line_prefix(o), line);
1209 break;
1210 case DIFF_SYMBOL_SUBMODULE_MODIFIED:
1211 fprintf(o->file, "%sSubmodule %s contains modified content\n",
1212 diff_line_prefix(o), line);
1213 break;
1214 case DIFF_SYMBOL_STATS_SUMMARY_NO_FILES:
1215 emit_line(o, "", "", " 0 files changed\n",
1216 strlen(" 0 files changed\n"));
1217 break;
1218 case DIFF_SYMBOL_STATS_SUMMARY_ABBREV:
1219 emit_line(o, "", "", " ...\n", strlen(" ...\n"));
1220 break;
1221 case DIFF_SYMBOL_WORD_DIFF:
1222 fprintf(o->file, "%.*s", len, line);
1223 break;
1224 case DIFF_SYMBOL_STAT_SEP:
1225 fputs(o->stat_sep, o->file);
1226 break;
1227 default:
1228 die("BUG: unknown diff symbol");
1229 }
1230 strbuf_release(&sb);
1231 }
1232
1233 static void emit_diff_symbol(struct diff_options *o, enum diff_symbol s,
1234 const char *line, int len, unsigned flags)
1235 {
1236 struct emitted_diff_symbol e = {line, len, flags, s};
1237
1238 if (o->emitted_symbols)
1239 append_emitted_diff_symbol(o, &e);
1240 else
1241 emit_diff_symbol_from_struct(o, &e);
1242 }
1243
1244 void diff_emit_submodule_del(struct diff_options *o, const char *line)
1245 {
1246 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_DEL, line, strlen(line), 0);
1247 }
1248
1249 void diff_emit_submodule_add(struct diff_options *o, const char *line)
1250 {
1251 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_ADD, line, strlen(line), 0);
1252 }
1253
1254 void diff_emit_submodule_untracked(struct diff_options *o, const char *path)
1255 {
1256 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_UNTRACKED,
1257 path, strlen(path), 0);
1258 }
1259
1260 void diff_emit_submodule_modified(struct diff_options *o, const char *path)
1261 {
1262 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_MODIFIED,
1263 path, strlen(path), 0);
1264 }
1265
1266 void diff_emit_submodule_header(struct diff_options *o, const char *header)
1267 {
1268 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_HEADER,
1269 header, strlen(header), 0);
1270 }
1271
1272 void diff_emit_submodule_error(struct diff_options *o, const char *err)
1273 {
1274 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_ERROR, err, strlen(err), 0);
1275 }
1276
1277 void diff_emit_submodule_pipethrough(struct diff_options *o,
1278 const char *line, int len)
1279 {
1280 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_PIPETHROUGH, line, len, 0);
1281 }
1282
1283 static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
1284 {
1285 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
1286 ecbdata->blank_at_eof_in_preimage &&
1287 ecbdata->blank_at_eof_in_postimage &&
1288 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
1289 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
1290 return 0;
1291 return ws_blank_line(line, len, ecbdata->ws_rule);
1292 }
1293
1294 static void emit_add_line(const char *reset,
1295 struct emit_callback *ecbdata,
1296 const char *line, int len)
1297 {
1298 unsigned flags = WSEH_NEW | ecbdata->ws_rule;
1299 if (new_blank_line_at_eof(ecbdata, line, len))
1300 flags |= DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF;
1301
1302 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_PLUS, line, len, flags);
1303 }
1304
1305 static void emit_del_line(const char *reset,
1306 struct emit_callback *ecbdata,
1307 const char *line, int len)
1308 {
1309 unsigned flags = WSEH_OLD | ecbdata->ws_rule;
1310 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_MINUS, line, len, flags);
1311 }
1312
1313 static void emit_context_line(const char *reset,
1314 struct emit_callback *ecbdata,
1315 const char *line, int len)
1316 {
1317 unsigned flags = WSEH_CONTEXT | ecbdata->ws_rule;
1318 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_CONTEXT, line, len, flags);
1319 }
1320
1321 static void emit_hunk_header(struct emit_callback *ecbdata,
1322 const char *line, int len)
1323 {
1324 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
1325 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
1326 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
1327 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1328 static const char atat[2] = { '@', '@' };
1329 const char *cp, *ep;
1330 struct strbuf msgbuf = STRBUF_INIT;
1331 int org_len = len;
1332 int i = 1;
1333
1334 /*
1335 * As a hunk header must begin with "@@ -<old>, +<new> @@",
1336 * it always is at least 10 bytes long.
1337 */
1338 if (len < 10 ||
1339 memcmp(line, atat, 2) ||
1340 !(ep = memmem(line + 2, len - 2, atat, 2))) {
1341 emit_diff_symbol(ecbdata->opt,
1342 DIFF_SYMBOL_CONTEXT_MARKER, line, len, 0);
1343 return;
1344 }
1345 ep += 2; /* skip over @@ */
1346
1347 /* The hunk header in fraginfo color */
1348 strbuf_addstr(&msgbuf, frag);
1349 strbuf_add(&msgbuf, line, ep - line);
1350 strbuf_addstr(&msgbuf, reset);
1351
1352 /*
1353 * trailing "\r\n"
1354 */
1355 for ( ; i < 3; i++)
1356 if (line[len - i] == '\r' || line[len - i] == '\n')
1357 len--;
1358
1359 /* blank before the func header */
1360 for (cp = ep; ep - line < len; ep++)
1361 if (*ep != ' ' && *ep != '\t')
1362 break;
1363 if (ep != cp) {
1364 strbuf_addstr(&msgbuf, context);
1365 strbuf_add(&msgbuf, cp, ep - cp);
1366 strbuf_addstr(&msgbuf, reset);
1367 }
1368
1369 if (ep < line + len) {
1370 strbuf_addstr(&msgbuf, func);
1371 strbuf_add(&msgbuf, ep, line + len - ep);
1372 strbuf_addstr(&msgbuf, reset);
1373 }
1374
1375 strbuf_add(&msgbuf, line + len, org_len - len);
1376 strbuf_complete_line(&msgbuf);
1377 emit_diff_symbol(ecbdata->opt,
1378 DIFF_SYMBOL_CONTEXT_FRAGINFO, msgbuf.buf, msgbuf.len, 0);
1379 strbuf_release(&msgbuf);
1380 }
1381
1382 static struct diff_tempfile *claim_diff_tempfile(void) {
1383 int i;
1384 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
1385 if (!diff_temp[i].name)
1386 return diff_temp + i;
1387 die("BUG: diff is failing to clean up its tempfiles");
1388 }
1389
1390 static void remove_tempfile(void)
1391 {
1392 int i;
1393 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
1394 if (is_tempfile_active(&diff_temp[i].tempfile))
1395 delete_tempfile(&diff_temp[i].tempfile);
1396 diff_temp[i].name = NULL;
1397 }
1398 }
1399
1400 static void add_line_count(struct strbuf *out, int count)
1401 {
1402 switch (count) {
1403 case 0:
1404 strbuf_addstr(out, "0,0");
1405 break;
1406 case 1:
1407 strbuf_addstr(out, "1");
1408 break;
1409 default:
1410 strbuf_addf(out, "1,%d", count);
1411 break;
1412 }
1413 }
1414
1415 static void emit_rewrite_lines(struct emit_callback *ecb,
1416 int prefix, const char *data, int size)
1417 {
1418 const char *endp = NULL;
1419 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
1420
1421 while (0 < size) {
1422 int len;
1423
1424 endp = memchr(data, '\n', size);
1425 len = endp ? (endp - data + 1) : size;
1426 if (prefix != '+') {
1427 ecb->lno_in_preimage++;
1428 emit_del_line(reset, ecb, data, len);
1429 } else {
1430 ecb->lno_in_postimage++;
1431 emit_add_line(reset, ecb, data, len);
1432 }
1433 size -= len;
1434 data += len;
1435 }
1436 if (!endp)
1437 emit_diff_symbol(ecb->opt, DIFF_SYMBOL_NO_LF_EOF, NULL, 0, 0);
1438 }
1439
1440 static void emit_rewrite_diff(const char *name_a,
1441 const char *name_b,
1442 struct diff_filespec *one,
1443 struct diff_filespec *two,
1444 struct userdiff_driver *textconv_one,
1445 struct userdiff_driver *textconv_two,
1446 struct diff_options *o)
1447 {
1448 int lc_a, lc_b;
1449 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
1450 const char *a_prefix, *b_prefix;
1451 char *data_one, *data_two;
1452 size_t size_one, size_two;
1453 struct emit_callback ecbdata;
1454 struct strbuf out = STRBUF_INIT;
1455
1456 if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
1457 a_prefix = o->b_prefix;
1458 b_prefix = o->a_prefix;
1459 } else {
1460 a_prefix = o->a_prefix;
1461 b_prefix = o->b_prefix;
1462 }
1463
1464 name_a += (*name_a == '/');
1465 name_b += (*name_b == '/');
1466
1467 strbuf_reset(&a_name);
1468 strbuf_reset(&b_name);
1469 quote_two_c_style(&a_name, a_prefix, name_a, 0);
1470 quote_two_c_style(&b_name, b_prefix, name_b, 0);
1471
1472 size_one = fill_textconv(textconv_one, one, &data_one);
1473 size_two = fill_textconv(textconv_two, two, &data_two);
1474
1475 memset(&ecbdata, 0, sizeof(ecbdata));
1476 ecbdata.color_diff = want_color(o->use_color);
1477 ecbdata.ws_rule = whitespace_rule(name_b);
1478 ecbdata.opt = o;
1479 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
1480 mmfile_t mf1, mf2;
1481 mf1.ptr = (char *)data_one;
1482 mf2.ptr = (char *)data_two;
1483 mf1.size = size_one;
1484 mf2.size = size_two;
1485 check_blank_at_eof(&mf1, &mf2, &ecbdata);
1486 }
1487 ecbdata.lno_in_preimage = 1;
1488 ecbdata.lno_in_postimage = 1;
1489
1490 lc_a = count_lines(data_one, size_one);
1491 lc_b = count_lines(data_two, size_two);
1492
1493 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
1494 a_name.buf, a_name.len, 0);
1495 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
1496 b_name.buf, b_name.len, 0);
1497
1498 strbuf_addstr(&out, "@@ -");
1499 if (!o->irreversible_delete)
1500 add_line_count(&out, lc_a);
1501 else
1502 strbuf_addstr(&out, "?,?");
1503 strbuf_addstr(&out, " +");
1504 add_line_count(&out, lc_b);
1505 strbuf_addstr(&out, " @@\n");
1506 emit_diff_symbol(o, DIFF_SYMBOL_REWRITE_DIFF, out.buf, out.len, 0);
1507 strbuf_release(&out);
1508
1509 if (lc_a && !o->irreversible_delete)
1510 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
1511 if (lc_b)
1512 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
1513 if (textconv_one)
1514 free((char *)data_one);
1515 if (textconv_two)
1516 free((char *)data_two);
1517 }
1518
1519 struct diff_words_buffer {
1520 mmfile_t text;
1521 long alloc;
1522 struct diff_words_orig {
1523 const char *begin, *end;
1524 } *orig;
1525 int orig_nr, orig_alloc;
1526 };
1527
1528 static void diff_words_append(char *line, unsigned long len,
1529 struct diff_words_buffer *buffer)
1530 {
1531 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
1532 line++;
1533 len--;
1534 memcpy(buffer->text.ptr + buffer->text.size, line, len);
1535 buffer->text.size += len;
1536 buffer->text.ptr[buffer->text.size] = '\0';
1537 }
1538
1539 struct diff_words_style_elem {
1540 const char *prefix;
1541 const char *suffix;
1542 const char *color; /* NULL; filled in by the setup code if
1543 * color is enabled */
1544 };
1545
1546 struct diff_words_style {
1547 enum diff_words_type type;
1548 struct diff_words_style_elem new, old, ctx;
1549 const char *newline;
1550 };
1551
1552 static struct diff_words_style diff_words_styles[] = {
1553 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
1554 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
1555 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
1556 };
1557
1558 struct diff_words_data {
1559 struct diff_words_buffer minus, plus;
1560 const char *current_plus;
1561 int last_minus;
1562 struct diff_options *opt;
1563 regex_t *word_regex;
1564 enum diff_words_type type;
1565 struct diff_words_style *style;
1566 };
1567
1568 static int fn_out_diff_words_write_helper(struct diff_options *o,
1569 struct diff_words_style_elem *st_el,
1570 const char *newline,
1571 size_t count, const char *buf)
1572 {
1573 int print = 0;
1574 struct strbuf sb = STRBUF_INIT;
1575
1576 while (count) {
1577 char *p = memchr(buf, '\n', count);
1578 if (print)
1579 strbuf_addstr(&sb, diff_line_prefix(o));
1580
1581 if (p != buf) {
1582 const char *reset = st_el->color && *st_el->color ?
1583 GIT_COLOR_RESET : NULL;
1584 if (st_el->color && *st_el->color)
1585 strbuf_addstr(&sb, st_el->color);
1586 strbuf_addstr(&sb, st_el->prefix);
1587 strbuf_add(&sb, buf, p ? p - buf : count);
1588 strbuf_addstr(&sb, st_el->suffix);
1589 if (reset)
1590 strbuf_addstr(&sb, reset);
1591 }
1592 if (!p)
1593 goto out;
1594
1595 strbuf_addstr(&sb, newline);
1596 count -= p + 1 - buf;
1597 buf = p + 1;
1598 print = 1;
1599 if (count) {
1600 emit_diff_symbol(o, DIFF_SYMBOL_WORD_DIFF,
1601 sb.buf, sb.len, 0);
1602 strbuf_reset(&sb);
1603 }
1604 }
1605
1606 out:
1607 if (sb.len)
1608 emit_diff_symbol(o, DIFF_SYMBOL_WORD_DIFF,
1609 sb.buf, sb.len, 0);
1610 strbuf_release(&sb);
1611 return 0;
1612 }
1613
1614 /*
1615 * '--color-words' algorithm can be described as:
1616 *
1617 * 1. collect the minus/plus lines of a diff hunk, divided into
1618 * minus-lines and plus-lines;
1619 *
1620 * 2. break both minus-lines and plus-lines into words and
1621 * place them into two mmfile_t with one word for each line;
1622 *
1623 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
1624 *
1625 * And for the common parts of the both file, we output the plus side text.
1626 * diff_words->current_plus is used to trace the current position of the plus file
1627 * which printed. diff_words->last_minus is used to trace the last minus word
1628 * printed.
1629 *
1630 * For '--graph' to work with '--color-words', we need to output the graph prefix
1631 * on each line of color words output. Generally, there are two conditions on
1632 * which we should output the prefix.
1633 *
1634 * 1. diff_words->last_minus == 0 &&
1635 * diff_words->current_plus == diff_words->plus.text.ptr
1636 *
1637 * that is: the plus text must start as a new line, and if there is no minus
1638 * word printed, a graph prefix must be printed.
1639 *
1640 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
1641 * *(diff_words->current_plus - 1) == '\n'
1642 *
1643 * that is: a graph prefix must be printed following a '\n'
1644 */
1645 static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
1646 {
1647 if ((diff_words->last_minus == 0 &&
1648 diff_words->current_plus == diff_words->plus.text.ptr) ||
1649 (diff_words->current_plus > diff_words->plus.text.ptr &&
1650 *(diff_words->current_plus - 1) == '\n')) {
1651 return 1;
1652 } else {
1653 return 0;
1654 }
1655 }
1656
1657 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
1658 {
1659 struct diff_words_data *diff_words = priv;
1660 struct diff_words_style *style = diff_words->style;
1661 int minus_first, minus_len, plus_first, plus_len;
1662 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
1663 struct diff_options *opt = diff_words->opt;
1664 const char *line_prefix;
1665
1666 if (line[0] != '@' || parse_hunk_header(line, len,
1667 &minus_first, &minus_len, &plus_first, &plus_len))
1668 return;
1669
1670 assert(opt);
1671 line_prefix = diff_line_prefix(opt);
1672
1673 /* POSIX requires that first be decremented by one if len == 0... */
1674 if (minus_len) {
1675 minus_begin = diff_words->minus.orig[minus_first].begin;
1676 minus_end =
1677 diff_words->minus.orig[minus_first + minus_len - 1].end;
1678 } else
1679 minus_begin = minus_end =
1680 diff_words->minus.orig[minus_first].end;
1681
1682 if (plus_len) {
1683 plus_begin = diff_words->plus.orig[plus_first].begin;
1684 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
1685 } else
1686 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
1687
1688 if (color_words_output_graph_prefix(diff_words)) {
1689 fputs(line_prefix, diff_words->opt->file);
1690 }
1691 if (diff_words->current_plus != plus_begin) {
1692 fn_out_diff_words_write_helper(diff_words->opt,
1693 &style->ctx, style->newline,
1694 plus_begin - diff_words->current_plus,
1695 diff_words->current_plus);
1696 }
1697 if (minus_begin != minus_end) {
1698 fn_out_diff_words_write_helper(diff_words->opt,
1699 &style->old, style->newline,
1700 minus_end - minus_begin, minus_begin);
1701 }
1702 if (plus_begin != plus_end) {
1703 fn_out_diff_words_write_helper(diff_words->opt,
1704 &style->new, style->newline,
1705 plus_end - plus_begin, plus_begin);
1706 }
1707
1708 diff_words->current_plus = plus_end;
1709 diff_words->last_minus = minus_first;
1710 }
1711
1712 /* This function starts looking at *begin, and returns 0 iff a word was found. */
1713 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
1714 int *begin, int *end)
1715 {
1716 if (word_regex && *begin < buffer->size) {
1717 regmatch_t match[1];
1718 if (!regexec_buf(word_regex, buffer->ptr + *begin,
1719 buffer->size - *begin, 1, match, 0)) {
1720 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
1721 '\n', match[0].rm_eo - match[0].rm_so);
1722 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
1723 *begin += match[0].rm_so;
1724 return *begin >= *end;
1725 }
1726 return -1;
1727 }
1728
1729 /* find the next word */
1730 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
1731 (*begin)++;
1732 if (*begin >= buffer->size)
1733 return -1;
1734
1735 /* find the end of the word */
1736 *end = *begin + 1;
1737 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
1738 (*end)++;
1739
1740 return 0;
1741 }
1742
1743 /*
1744 * This function splits the words in buffer->text, stores the list with
1745 * newline separator into out, and saves the offsets of the original words
1746 * in buffer->orig.
1747 */
1748 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
1749 regex_t *word_regex)
1750 {
1751 int i, j;
1752 long alloc = 0;
1753
1754 out->size = 0;
1755 out->ptr = NULL;
1756
1757 /* fake an empty "0th" word */
1758 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
1759 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
1760 buffer->orig_nr = 1;
1761
1762 for (i = 0; i < buffer->text.size; i++) {
1763 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
1764 return;
1765
1766 /* store original boundaries */
1767 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
1768 buffer->orig_alloc);
1769 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
1770 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
1771 buffer->orig_nr++;
1772
1773 /* store one word */
1774 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
1775 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
1776 out->ptr[out->size + j - i] = '\n';
1777 out->size += j - i + 1;
1778
1779 i = j - 1;
1780 }
1781 }
1782
1783 /* this executes the word diff on the accumulated buffers */
1784 static void diff_words_show(struct diff_words_data *diff_words)
1785 {
1786 xpparam_t xpp;
1787 xdemitconf_t xecfg;
1788 mmfile_t minus, plus;
1789 struct diff_words_style *style = diff_words->style;
1790
1791 struct diff_options *opt = diff_words->opt;
1792 const char *line_prefix;
1793
1794 assert(opt);
1795 line_prefix = diff_line_prefix(opt);
1796
1797 /* special case: only removal */
1798 if (!diff_words->plus.text.size) {
1799 emit_diff_symbol(diff_words->opt, DIFF_SYMBOL_WORD_DIFF,
1800 line_prefix, strlen(line_prefix), 0);
1801 fn_out_diff_words_write_helper(diff_words->opt,
1802 &style->old, style->newline,
1803 diff_words->minus.text.size,
1804 diff_words->minus.text.ptr);
1805 diff_words->minus.text.size = 0;
1806 return;
1807 }
1808
1809 diff_words->current_plus = diff_words->plus.text.ptr;
1810 diff_words->last_minus = 0;
1811
1812 memset(&xpp, 0, sizeof(xpp));
1813 memset(&xecfg, 0, sizeof(xecfg));
1814 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1815 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1816 xpp.flags = 0;
1817 /* as only the hunk header will be parsed, we need a 0-context */
1818 xecfg.ctxlen = 0;
1819 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1820 &xpp, &xecfg))
1821 die("unable to generate word diff");
1822 free(minus.ptr);
1823 free(plus.ptr);
1824 if (diff_words->current_plus != diff_words->plus.text.ptr +
1825 diff_words->plus.text.size) {
1826 if (color_words_output_graph_prefix(diff_words))
1827 emit_diff_symbol(diff_words->opt, DIFF_SYMBOL_WORD_DIFF,
1828 line_prefix, strlen(line_prefix), 0);
1829 fn_out_diff_words_write_helper(diff_words->opt,
1830 &style->ctx, style->newline,
1831 diff_words->plus.text.ptr + diff_words->plus.text.size
1832 - diff_words->current_plus, diff_words->current_plus);
1833 }
1834 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1835 }
1836
1837 /* In "color-words" mode, show word-diff of words accumulated in the buffer */
1838 static void diff_words_flush(struct emit_callback *ecbdata)
1839 {
1840 struct diff_options *wo = ecbdata->diff_words->opt;
1841
1842 if (ecbdata->diff_words->minus.text.size ||
1843 ecbdata->diff_words->plus.text.size)
1844 diff_words_show(ecbdata->diff_words);
1845
1846 if (wo->emitted_symbols) {
1847 struct diff_options *o = ecbdata->opt;
1848 struct emitted_diff_symbols *wol = wo->emitted_symbols;
1849 int i;
1850
1851 /*
1852 * NEEDSWORK:
1853 * Instead of appending each, concat all words to a line?
1854 */
1855 for (i = 0; i < wol->nr; i++)
1856 append_emitted_diff_symbol(o, &wol->buf[i]);
1857
1858 for (i = 0; i < wol->nr; i++)
1859 free((void *)wol->buf[i].line);
1860
1861 wol->nr = 0;
1862 }
1863 }
1864
1865 static void diff_filespec_load_driver(struct diff_filespec *one)
1866 {
1867 /* Use already-loaded driver */
1868 if (one->driver)
1869 return;
1870
1871 if (S_ISREG(one->mode))
1872 one->driver = userdiff_find_by_path(one->path);
1873
1874 /* Fallback to default settings */
1875 if (!one->driver)
1876 one->driver = userdiff_find_by_name("default");
1877 }
1878
1879 static const char *userdiff_word_regex(struct diff_filespec *one)
1880 {
1881 diff_filespec_load_driver(one);
1882 return one->driver->word_regex;
1883 }
1884
1885 static void init_diff_words_data(struct emit_callback *ecbdata,
1886 struct diff_options *orig_opts,
1887 struct diff_filespec *one,
1888 struct diff_filespec *two)
1889 {
1890 int i;
1891 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1892 memcpy(o, orig_opts, sizeof(struct diff_options));
1893
1894 ecbdata->diff_words =
1895 xcalloc(1, sizeof(struct diff_words_data));
1896 ecbdata->diff_words->type = o->word_diff;
1897 ecbdata->diff_words->opt = o;
1898
1899 if (orig_opts->emitted_symbols)
1900 o->emitted_symbols =
1901 xcalloc(1, sizeof(struct emitted_diff_symbols));
1902
1903 if (!o->word_regex)
1904 o->word_regex = userdiff_word_regex(one);
1905 if (!o->word_regex)
1906 o->word_regex = userdiff_word_regex(two);
1907 if (!o->word_regex)
1908 o->word_regex = diff_word_regex_cfg;
1909 if (o->word_regex) {
1910 ecbdata->diff_words->word_regex = (regex_t *)
1911 xmalloc(sizeof(regex_t));
1912 if (regcomp(ecbdata->diff_words->word_regex,
1913 o->word_regex,
1914 REG_EXTENDED | REG_NEWLINE))
1915 die ("Invalid regular expression: %s",
1916 o->word_regex);
1917 }
1918 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1919 if (o->word_diff == diff_words_styles[i].type) {
1920 ecbdata->diff_words->style =
1921 &diff_words_styles[i];
1922 break;
1923 }
1924 }
1925 if (want_color(o->use_color)) {
1926 struct diff_words_style *st = ecbdata->diff_words->style;
1927 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1928 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1929 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
1930 }
1931 }
1932
1933 static void free_diff_words_data(struct emit_callback *ecbdata)
1934 {
1935 if (ecbdata->diff_words) {
1936 diff_words_flush(ecbdata);
1937 free (ecbdata->diff_words->opt->emitted_symbols);
1938 free (ecbdata->diff_words->opt);
1939 free (ecbdata->diff_words->minus.text.ptr);
1940 free (ecbdata->diff_words->minus.orig);
1941 free (ecbdata->diff_words->plus.text.ptr);
1942 free (ecbdata->diff_words->plus.orig);
1943 if (ecbdata->diff_words->word_regex) {
1944 regfree(ecbdata->diff_words->word_regex);
1945 free(ecbdata->diff_words->word_regex);
1946 }
1947 FREE_AND_NULL(ecbdata->diff_words);
1948 }
1949 }
1950
1951 const char *diff_get_color(int diff_use_color, enum color_diff ix)
1952 {
1953 if (want_color(diff_use_color))
1954 return diff_colors[ix];
1955 return "";
1956 }
1957
1958 const char *diff_line_prefix(struct diff_options *opt)
1959 {
1960 struct strbuf *msgbuf;
1961 if (!opt->output_prefix)
1962 return "";
1963
1964 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1965 return msgbuf->buf;
1966 }
1967
1968 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1969 {
1970 const char *cp;
1971 unsigned long allot;
1972 size_t l = len;
1973
1974 if (ecb->truncate)
1975 return ecb->truncate(line, len);
1976 cp = line;
1977 allot = l;
1978 while (0 < l) {
1979 (void) utf8_width(&cp, &l);
1980 if (!cp)
1981 break; /* truncated in the middle? */
1982 }
1983 return allot - l;
1984 }
1985
1986 static void find_lno(const char *line, struct emit_callback *ecbdata)
1987 {
1988 const char *p;
1989 ecbdata->lno_in_preimage = 0;
1990 ecbdata->lno_in_postimage = 0;
1991 p = strchr(line, '-');
1992 if (!p)
1993 return; /* cannot happen */
1994 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1995 p = strchr(p, '+');
1996 if (!p)
1997 return; /* cannot happen */
1998 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1999 }
2000
2001 static void fn_out_consume(void *priv, char *line, unsigned long len)
2002 {
2003 struct emit_callback *ecbdata = priv;
2004 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
2005 struct diff_options *o = ecbdata->opt;
2006
2007 o->found_changes = 1;
2008
2009 if (ecbdata->header) {
2010 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
2011 ecbdata->header->buf, ecbdata->header->len, 0);
2012 strbuf_reset(ecbdata->header);
2013 ecbdata->header = NULL;
2014 }
2015
2016 if (ecbdata->label_path[0]) {
2017 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
2018 ecbdata->label_path[0],
2019 strlen(ecbdata->label_path[0]), 0);
2020 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
2021 ecbdata->label_path[1],
2022 strlen(ecbdata->label_path[1]), 0);
2023 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
2024 }
2025
2026 if (diff_suppress_blank_empty
2027 && len == 2 && line[0] == ' ' && line[1] == '\n') {
2028 line[0] = '\n';
2029 len = 1;
2030 }
2031
2032 if (line[0] == '@') {
2033 if (ecbdata->diff_words)
2034 diff_words_flush(ecbdata);
2035 len = sane_truncate_line(ecbdata, line, len);
2036 find_lno(line, ecbdata);
2037 emit_hunk_header(ecbdata, line, len);
2038 return;
2039 }
2040
2041 if (ecbdata->diff_words) {
2042 enum diff_symbol s =
2043 ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN ?
2044 DIFF_SYMBOL_WORDS_PORCELAIN : DIFF_SYMBOL_WORDS;
2045 if (line[0] == '-') {
2046 diff_words_append(line, len,
2047 &ecbdata->diff_words->minus);
2048 return;
2049 } else if (line[0] == '+') {
2050 diff_words_append(line, len,
2051 &ecbdata->diff_words->plus);
2052 return;
2053 } else if (starts_with(line, "\\ ")) {
2054 /*
2055 * Eat the "no newline at eof" marker as if we
2056 * saw a "+" or "-" line with nothing on it,
2057 * and return without diff_words_flush() to
2058 * defer processing. If this is the end of
2059 * preimage, more "+" lines may come after it.
2060 */
2061 return;
2062 }
2063 diff_words_flush(ecbdata);
2064 emit_diff_symbol(o, s, line, len, 0);
2065 return;
2066 }
2067
2068 switch (line[0]) {
2069 case '+':
2070 ecbdata->lno_in_postimage++;
2071 emit_add_line(reset, ecbdata, line + 1, len - 1);
2072 break;
2073 case '-':
2074 ecbdata->lno_in_preimage++;
2075 emit_del_line(reset, ecbdata, line + 1, len - 1);
2076 break;
2077 case ' ':
2078 ecbdata->lno_in_postimage++;
2079 ecbdata->lno_in_preimage++;
2080 emit_context_line(reset, ecbdata, line + 1, len - 1);
2081 break;
2082 default:
2083 /* incomplete line at the end */
2084 ecbdata->lno_in_preimage++;
2085 emit_diff_symbol(o, DIFF_SYMBOL_CONTEXT_INCOMPLETE,
2086 line, len, 0);
2087 break;
2088 }
2089 }
2090
2091 static char *pprint_rename(const char *a, const char *b)
2092 {
2093 const char *old = a;
2094 const char *new = b;
2095 struct strbuf name = STRBUF_INIT;
2096 int pfx_length, sfx_length;
2097 int pfx_adjust_for_slash;
2098 int len_a = strlen(a);
2099 int len_b = strlen(b);
2100 int a_midlen, b_midlen;
2101 int qlen_a = quote_c_style(a, NULL, NULL, 0);
2102 int qlen_b = quote_c_style(b, NULL, NULL, 0);
2103
2104 if (qlen_a || qlen_b) {
2105 quote_c_style(a, &name, NULL, 0);
2106 strbuf_addstr(&name, " => ");
2107 quote_c_style(b, &name, NULL, 0);
2108 return strbuf_detach(&name, NULL);
2109 }
2110
2111 /* Find common prefix */
2112 pfx_length = 0;
2113 while (*old && *new && *old == *new) {
2114 if (*old == '/')
2115 pfx_length = old - a + 1;
2116 old++;
2117 new++;
2118 }
2119
2120 /* Find common suffix */
2121 old = a + len_a;
2122 new = b + len_b;
2123 sfx_length = 0;
2124 /*
2125 * If there is a common prefix, it must end in a slash. In
2126 * that case we let this loop run 1 into the prefix to see the
2127 * same slash.
2128 *
2129 * If there is no common prefix, we cannot do this as it would
2130 * underrun the input strings.
2131 */
2132 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
2133 while (a + pfx_length - pfx_adjust_for_slash <= old &&
2134 b + pfx_length - pfx_adjust_for_slash <= new &&
2135 *old == *new) {
2136 if (*old == '/')
2137 sfx_length = len_a - (old - a);
2138 old--;
2139 new--;
2140 }
2141
2142 /*
2143 * pfx{mid-a => mid-b}sfx
2144 * {pfx-a => pfx-b}sfx
2145 * pfx{sfx-a => sfx-b}
2146 * name-a => name-b
2147 */
2148 a_midlen = len_a - pfx_length - sfx_length;
2149 b_midlen = len_b - pfx_length - sfx_length;
2150 if (a_midlen < 0)
2151 a_midlen = 0;
2152 if (b_midlen < 0)
2153 b_midlen = 0;
2154
2155 strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
2156 if (pfx_length + sfx_length) {
2157 strbuf_add(&name, a, pfx_length);
2158 strbuf_addch(&name, '{');
2159 }
2160 strbuf_add(&name, a + pfx_length, a_midlen);
2161 strbuf_addstr(&name, " => ");
2162 strbuf_add(&name, b + pfx_length, b_midlen);
2163 if (pfx_length + sfx_length) {
2164 strbuf_addch(&name, '}');
2165 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
2166 }
2167 return strbuf_detach(&name, NULL);
2168 }
2169
2170 struct diffstat_t {
2171 int nr;
2172 int alloc;
2173 struct diffstat_file {
2174 char *from_name;
2175 char *name;
2176 char *print_name;
2177 unsigned is_unmerged:1;
2178 unsigned is_binary:1;
2179 unsigned is_renamed:1;
2180 unsigned is_interesting:1;
2181 uintmax_t added, deleted;
2182 } **files;
2183 };
2184
2185 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
2186 const char *name_a,
2187 const char *name_b)
2188 {
2189 struct diffstat_file *x;
2190 x = xcalloc(1, sizeof(*x));
2191 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
2192 diffstat->files[diffstat->nr++] = x;
2193 if (name_b) {
2194 x->from_name = xstrdup(name_a);
2195 x->name = xstrdup(name_b);
2196 x->is_renamed = 1;
2197 }
2198 else {
2199 x->from_name = NULL;
2200 x->name = xstrdup(name_a);
2201 }
2202 return x;
2203 }
2204
2205 static void diffstat_consume(void *priv, char *line, unsigned long len)
2206 {
2207 struct diffstat_t *diffstat = priv;
2208 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
2209
2210 if (line[0] == '+')
2211 x->added++;
2212 else if (line[0] == '-')
2213 x->deleted++;
2214 }
2215
2216 const char mime_boundary_leader[] = "------------";
2217
2218 static int scale_linear(int it, int width, int max_change)
2219 {
2220 if (!it)
2221 return 0;
2222 /*
2223 * make sure that at least one '-' or '+' is printed if
2224 * there is any change to this path. The easiest way is to
2225 * scale linearly as if the alloted width is one column shorter
2226 * than it is, and then add 1 to the result.
2227 */
2228 return 1 + (it * (width - 1) / max_change);
2229 }
2230
2231 static void show_graph(struct strbuf *out, char ch, int cnt,
2232 const char *set, const char *reset)
2233 {
2234 if (cnt <= 0)
2235 return;
2236 strbuf_addstr(out, set);
2237 strbuf_addchars(out, ch, cnt);
2238 strbuf_addstr(out, reset);
2239 }
2240
2241 static void fill_print_name(struct diffstat_file *file)
2242 {
2243 char *pname;
2244
2245 if (file->print_name)
2246 return;
2247
2248 if (!file->is_renamed) {
2249 struct strbuf buf = STRBUF_INIT;
2250 if (quote_c_style(file->name, &buf, NULL, 0)) {
2251 pname = strbuf_detach(&buf, NULL);
2252 } else {
2253 pname = file->name;
2254 strbuf_release(&buf);
2255 }
2256 } else {
2257 pname = pprint_rename(file->from_name, file->name);
2258 }
2259 file->print_name = pname;
2260 }
2261
2262 static void print_stat_summary_inserts_deletes(struct diff_options *options,
2263 int files, int insertions, int deletions)
2264 {
2265 struct strbuf sb = STRBUF_INIT;
2266
2267 if (!files) {
2268 assert(insertions == 0 && deletions == 0);
2269 emit_diff_symbol(options, DIFF_SYMBOL_STATS_SUMMARY_NO_FILES,
2270 NULL, 0, 0);
2271 return;
2272 }
2273
2274 strbuf_addf(&sb,
2275 (files == 1) ? " %d file changed" : " %d files changed",
2276 files);
2277
2278 /*
2279 * For binary diff, the caller may want to print "x files
2280 * changed" with insertions == 0 && deletions == 0.
2281 *
2282 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
2283 * is probably less confusing (i.e skip over "2 files changed
2284 * but nothing about added/removed lines? Is this a bug in Git?").
2285 */
2286 if (insertions || deletions == 0) {
2287 strbuf_addf(&sb,
2288 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
2289 insertions);
2290 }
2291
2292 if (deletions || insertions == 0) {
2293 strbuf_addf(&sb,
2294 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
2295 deletions);
2296 }
2297 strbuf_addch(&sb, '\n');
2298 emit_diff_symbol(options, DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES,
2299 sb.buf, sb.len, 0);
2300 strbuf_release(&sb);
2301 }
2302
2303 void print_stat_summary(FILE *fp, int files,
2304 int insertions, int deletions)
2305 {
2306 struct diff_options o;
2307 memset(&o, 0, sizeof(o));
2308 o.file = fp;
2309
2310 print_stat_summary_inserts_deletes(&o, files, insertions, deletions);
2311 }
2312
2313 static void show_stats(struct diffstat_t *data, struct diff_options *options)
2314 {
2315 int i, len, add, del, adds = 0, dels = 0;
2316 uintmax_t max_change = 0, max_len = 0;
2317 int total_files = data->nr, count;
2318 int width, name_width, graph_width, number_width = 0, bin_width = 0;
2319 const char *reset, *add_c, *del_c;
2320 int extra_shown = 0;
2321 const char *line_prefix = diff_line_prefix(options);
2322 struct strbuf out = STRBUF_INIT;
2323
2324 if (data->nr == 0)
2325 return;
2326
2327 count = options->stat_count ? options->stat_count : data->nr;
2328
2329 reset = diff_get_color_opt(options, DIFF_RESET);
2330 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
2331 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
2332
2333 /*
2334 * Find the longest filename and max number of changes
2335 */
2336 for (i = 0; (i < count) && (i < data->nr); i++) {
2337 struct diffstat_file *file = data->files[i];
2338 uintmax_t change = file->added + file->deleted;
2339
2340 if (!file->is_interesting && (change == 0)) {
2341 count++; /* not shown == room for one more */
2342 continue;
2343 }
2344 fill_print_name(file);
2345 len = strlen(file->print_name);
2346 if (max_len < len)
2347 max_len = len;
2348
2349 if (file->is_unmerged) {
2350 /* "Unmerged" is 8 characters */
2351 bin_width = bin_width < 8 ? 8 : bin_width;
2352 continue;
2353 }
2354 if (file->is_binary) {
2355 /* "Bin XXX -> YYY bytes" */
2356 int w = 14 + decimal_width(file->added)
2357 + decimal_width(file->deleted);
2358 bin_width = bin_width < w ? w : bin_width;
2359 /* Display change counts aligned with "Bin" */
2360 number_width = 3;
2361 continue;
2362 }
2363
2364 if (max_change < change)
2365 max_change = change;
2366 }
2367 count = i; /* where we can stop scanning in data->files[] */
2368
2369 /*
2370 * We have width = stat_width or term_columns() columns total.
2371 * We want a maximum of min(max_len, stat_name_width) for the name part.
2372 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
2373 * We also need 1 for " " and 4 + decimal_width(max_change)
2374 * for " | NNNN " and one the empty column at the end, altogether
2375 * 6 + decimal_width(max_change).
2376 *
2377 * If there's not enough space, we will use the smaller of
2378 * stat_name_width (if set) and 5/8*width for the filename,
2379 * and the rest for constant elements + graph part, but no more
2380 * than stat_graph_width for the graph part.
2381 * (5/8 gives 50 for filename and 30 for the constant parts + graph
2382 * for the standard terminal size).
2383 *
2384 * In other words: stat_width limits the maximum width, and
2385 * stat_name_width fixes the maximum width of the filename,
2386 * and is also used to divide available columns if there
2387 * aren't enough.
2388 *
2389 * Binary files are displayed with "Bin XXX -> YYY bytes"
2390 * instead of the change count and graph. This part is treated
2391 * similarly to the graph part, except that it is not
2392 * "scaled". If total width is too small to accommodate the
2393 * guaranteed minimum width of the filename part and the
2394 * separators and this message, this message will "overflow"
2395 * making the line longer than the maximum width.
2396 */
2397
2398 if (options->stat_width == -1)
2399 width = term_columns() - strlen(line_prefix);
2400 else
2401 width = options->stat_width ? options->stat_width : 80;
2402 number_width = decimal_width(max_change) > number_width ?
2403 decimal_width(max_change) : number_width;
2404
2405 if (options->stat_graph_width == -1)
2406 options->stat_graph_width = diff_stat_graph_width;
2407
2408 /*
2409 * Guarantee 3/8*16==6 for the graph part
2410 * and 5/8*16==10 for the filename part
2411 */
2412 if (width < 16 + 6 + number_width)
2413 width = 16 + 6 + number_width;
2414
2415 /*
2416 * First assign sizes that are wanted, ignoring available width.
2417 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
2418 * starting from "XXX" should fit in graph_width.
2419 */
2420 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
2421 if (options->stat_graph_width &&
2422 options->stat_graph_width < graph_width)
2423 graph_width = options->stat_graph_width;
2424
2425 name_width = (options->stat_name_width > 0 &&
2426 options->stat_name_width < max_len) ?
2427 options->stat_name_width : max_len;
2428
2429 /*
2430 * Adjust adjustable widths not to exceed maximum width
2431 */
2432 if (name_width + number_width + 6 + graph_width > width) {
2433 if (graph_width > width * 3/8 - number_width - 6) {
2434 graph_width = width * 3/8 - number_width - 6;
2435 if (graph_width < 6)
2436 graph_width = 6;
2437 }
2438
2439 if (options->stat_graph_width &&
2440 graph_width > options->stat_graph_width)
2441 graph_width = options->stat_graph_width;
2442 if (name_width > width - number_width - 6 - graph_width)
2443 name_width = width - number_width - 6 - graph_width;
2444 else
2445 graph_width = width - number_width - 6 - name_width;
2446 }
2447
2448 /*
2449 * From here name_width is the width of the name area,
2450 * and graph_width is the width of the graph area.
2451 * max_change is used to scale graph properly.
2452 */
2453 for (i = 0; i < count; i++) {
2454 const char *prefix = "";
2455 struct diffstat_file *file = data->files[i];
2456 char *name = file->print_name;
2457 uintmax_t added = file->added;
2458 uintmax_t deleted = file->deleted;
2459 int name_len;
2460
2461 if (!file->is_interesting && (added + deleted == 0))
2462 continue;
2463
2464 /*
2465 * "scale" the filename
2466 */
2467 len = name_width;
2468 name_len = strlen(name);
2469 if (name_width < name_len) {
2470 char *slash;
2471 prefix = "...";
2472 len -= 3;
2473 name += name_len - len;
2474 slash = strchr(name, '/');
2475 if (slash)
2476 name = slash;
2477 }
2478
2479 if (file->is_binary) {
2480 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2481 strbuf_addf(&out, " %*s", number_width, "Bin");
2482 if (!added && !deleted) {
2483 strbuf_addch(&out, '\n');
2484 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2485 out.buf, out.len, 0);
2486 strbuf_reset(&out);
2487 continue;
2488 }
2489 strbuf_addf(&out, " %s%"PRIuMAX"%s",
2490 del_c, deleted, reset);
2491 strbuf_addstr(&out, " -> ");
2492 strbuf_addf(&out, "%s%"PRIuMAX"%s",
2493 add_c, added, reset);
2494 strbuf_addstr(&out, " bytes\n");
2495 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2496 out.buf, out.len, 0);
2497 strbuf_reset(&out);
2498 continue;
2499 }
2500 else if (file->is_unmerged) {
2501 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2502 strbuf_addstr(&out, " Unmerged\n");
2503 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2504 out.buf, out.len, 0);
2505 strbuf_reset(&out);
2506 continue;
2507 }
2508
2509 /*
2510 * scale the add/delete
2511 */
2512 add = added;
2513 del = deleted;
2514
2515 if (graph_width <= max_change) {
2516 int total = scale_linear(add + del, graph_width, max_change);
2517 if (total < 2 && add && del)
2518 /* width >= 2 due to the sanity check */
2519 total = 2;
2520 if (add < del) {
2521 add = scale_linear(add, graph_width, max_change);
2522 del = total - add;
2523 } else {
2524 del = scale_linear(del, graph_width, max_change);
2525 add = total - del;
2526 }
2527 }
2528 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2529 strbuf_addf(&out, " %*"PRIuMAX"%s",
2530 number_width, added + deleted,
2531 added + deleted ? " " : "");
2532 show_graph(&out, '+', add, add_c, reset);
2533 show_graph(&out, '-', del, del_c, reset);
2534 strbuf_addch(&out, '\n');
2535 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2536 out.buf, out.len, 0);
2537 strbuf_reset(&out);
2538 }
2539
2540 for (i = 0; i < data->nr; i++) {
2541 struct diffstat_file *file = data->files[i];
2542 uintmax_t added = file->added;
2543 uintmax_t deleted = file->deleted;
2544
2545 if (file->is_unmerged ||
2546 (!file->is_interesting && (added + deleted == 0))) {
2547 total_files--;
2548 continue;
2549 }
2550
2551 if (!file->is_binary) {
2552 adds += added;
2553 dels += deleted;
2554 }
2555 if (i < count)
2556 continue;
2557 if (!extra_shown)
2558 emit_diff_symbol(options,
2559 DIFF_SYMBOL_STATS_SUMMARY_ABBREV,
2560 NULL, 0, 0);
2561 extra_shown = 1;
2562 }
2563
2564 print_stat_summary_inserts_deletes(options, total_files, adds, dels);
2565 }
2566
2567 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
2568 {
2569 int i, adds = 0, dels = 0, total_files = data->nr;
2570
2571 if (data->nr == 0)
2572 return;
2573
2574 for (i = 0; i < data->nr; i++) {
2575 int added = data->files[i]->added;
2576 int deleted = data->files[i]->deleted;
2577
2578 if (data->files[i]->is_unmerged ||
2579 (!data->files[i]->is_interesting && (added + deleted == 0))) {
2580 total_files--;
2581 } else if (!data->files[i]->is_binary) { /* don't count bytes */
2582 adds += added;
2583 dels += deleted;
2584 }
2585 }
2586 print_stat_summary_inserts_deletes(options, total_files, adds, dels);
2587 }
2588
2589 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
2590 {
2591 int i;
2592
2593 if (data->nr == 0)
2594 return;
2595
2596 for (i = 0; i < data->nr; i++) {
2597 struct diffstat_file *file = data->files[i];
2598
2599 fprintf(options->file, "%s", diff_line_prefix(options));
2600
2601 if (file->is_binary)
2602 fprintf(options->file, "-\t-\t");
2603 else
2604 fprintf(options->file,
2605 "%"PRIuMAX"\t%"PRIuMAX"\t",
2606 file->added, file->deleted);
2607 if (options->line_termination) {
2608 fill_print_name(file);
2609 if (!file->is_renamed)
2610 write_name_quoted(file->name, options->file,
2611 options->line_termination);
2612 else {
2613 fputs(file->print_name, options->file);
2614 putc(options->line_termination, options->file);
2615 }
2616 } else {
2617 if (file->is_renamed) {
2618 putc('\0', options->file);
2619 write_name_quoted(file->from_name, options->file, '\0');
2620 }
2621 write_name_quoted(file->name, options->file, '\0');
2622 }
2623 }
2624 }
2625
2626 struct dirstat_file {
2627 const char *name;
2628 unsigned long changed;
2629 };
2630
2631 struct dirstat_dir {
2632 struct dirstat_file *files;
2633 int alloc, nr, permille, cumulative;
2634 };
2635
2636 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
2637 unsigned long changed, const char *base, int baselen)
2638 {
2639 unsigned long this_dir = 0;
2640 unsigned int sources = 0;
2641 const char *line_prefix = diff_line_prefix(opt);
2642
2643 while (dir->nr) {
2644 struct dirstat_file *f = dir->files;
2645 int namelen = strlen(f->name);
2646 unsigned long this;
2647 char *slash;
2648
2649 if (namelen < baselen)
2650 break;
2651 if (memcmp(f->name, base, baselen))
2652 break;
2653 slash = strchr(f->name + baselen, '/');
2654 if (slash) {
2655 int newbaselen = slash + 1 - f->name;
2656 this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
2657 sources++;
2658 } else {
2659 this = f->changed;
2660 dir->files++;
2661 dir->nr--;
2662 sources += 2;
2663 }
2664 this_dir += this;
2665 }
2666
2667 /*
2668 * We don't report dirstat's for
2669 * - the top level
2670 * - or cases where everything came from a single directory
2671 * under this directory (sources == 1).
2672 */
2673 if (baselen && sources != 1) {
2674 if (this_dir) {
2675 int permille = this_dir * 1000 / changed;
2676 if (permille >= dir->permille) {
2677 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
2678 permille / 10, permille % 10, baselen, base);
2679 if (!dir->cumulative)
2680 return 0;
2681 }
2682 }
2683 }
2684 return this_dir;
2685 }
2686
2687 static int dirstat_compare(const void *_a, const void *_b)
2688 {
2689 const struct dirstat_file *a = _a;
2690 const struct dirstat_file *b = _b;
2691 return strcmp(a->name, b->name);
2692 }
2693
2694 static void show_dirstat(struct diff_options *options)
2695 {
2696 int i;
2697 unsigned long changed;
2698 struct dirstat_dir dir;
2699 struct diff_queue_struct *q = &diff_queued_diff;
2700
2701 dir.files = NULL;
2702 dir.alloc = 0;
2703 dir.nr = 0;
2704 dir.permille = options->dirstat_permille;
2705 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2706
2707 changed = 0;
2708 for (i = 0; i < q->nr; i++) {
2709 struct diff_filepair *p = q->queue[i];
2710 const char *name;
2711 unsigned long copied, added, damage;
2712 int content_changed;
2713
2714 name = p->two->path ? p->two->path : p->one->path;
2715
2716 if (p->one->oid_valid && p->two->oid_valid)
2717 content_changed = oidcmp(&p->one->oid, &p->two->oid);
2718 else
2719 content_changed = 1;
2720
2721 if (!content_changed) {
2722 /*
2723 * The SHA1 has not changed, so pre-/post-content is
2724 * identical. We can therefore skip looking at the
2725 * file contents altogether.
2726 */
2727 damage = 0;
2728 goto found_damage;
2729 }
2730
2731 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
2732 /*
2733 * In --dirstat-by-file mode, we don't really need to
2734 * look at the actual file contents at all.
2735 * The fact that the SHA1 changed is enough for us to
2736 * add this file to the list of results
2737 * (with each file contributing equal damage).
2738 */
2739 damage = 1;
2740 goto found_damage;
2741 }
2742
2743 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
2744 diff_populate_filespec(p->one, 0);
2745 diff_populate_filespec(p->two, 0);
2746 diffcore_count_changes(p->one, p->two, NULL, NULL,
2747 &copied, &added);
2748 diff_free_filespec_data(p->one);
2749 diff_free_filespec_data(p->two);
2750 } else if (DIFF_FILE_VALID(p->one)) {
2751 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
2752 copied = added = 0;
2753 diff_free_filespec_data(p->one);
2754 } else if (DIFF_FILE_VALID(p->two)) {
2755 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
2756 copied = 0;
2757 added = p->two->size;
2758 diff_free_filespec_data(p->two);
2759 } else
2760 continue;
2761
2762 /*
2763 * Original minus copied is the removed material,
2764 * added is the new material. They are both damages
2765 * made to the preimage.
2766 * If the resulting damage is zero, we know that
2767 * diffcore_count_changes() considers the two entries to
2768 * be identical, but since content_changed is true, we
2769 * know that there must have been _some_ kind of change,
2770 * so we force all entries to have damage > 0.
2771 */
2772 damage = (p->one->size - copied) + added;
2773 if (!damage)
2774 damage = 1;
2775
2776 found_damage:
2777 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2778 dir.files[dir.nr].name = name;
2779 dir.files[dir.nr].changed = damage;
2780 changed += damage;
2781 dir.nr++;
2782 }
2783
2784 /* This can happen even with many files, if everything was renames */
2785 if (!changed)
2786 return;
2787
2788 /* Show all directories with more than x% of the changes */
2789 QSORT(dir.files, dir.nr, dirstat_compare);
2790 gather_dirstat(options, &dir, changed, "", 0);
2791 }
2792
2793 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
2794 {
2795 int i;
2796 unsigned long changed;
2797 struct dirstat_dir dir;
2798
2799 if (data->nr == 0)
2800 return;
2801
2802 dir.files = NULL;
2803 dir.alloc = 0;
2804 dir.nr = 0;
2805 dir.permille = options->dirstat_permille;
2806 dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
2807
2808 changed = 0;
2809 for (i = 0; i < data->nr; i++) {
2810 struct diffstat_file *file = data->files[i];
2811 unsigned long damage = file->added + file->deleted;
2812 if (file->is_binary)
2813 /*
2814 * binary files counts bytes, not lines. Must find some
2815 * way to normalize binary bytes vs. textual lines.
2816 * The following heuristic assumes that there are 64
2817 * bytes per "line".
2818 * This is stupid and ugly, but very cheap...
2819 */
2820 damage = (damage + 63) / 64;
2821 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2822 dir.files[dir.nr].name = file->name;
2823 dir.files[dir.nr].changed = damage;
2824 changed += damage;
2825 dir.nr++;
2826 }
2827
2828 /* This can happen even with many files, if everything was renames */
2829 if (!changed)
2830 return;
2831
2832 /* Show all directories with more than x% of the changes */
2833 QSORT(dir.files, dir.nr, dirstat_compare);
2834 gather_dirstat(options, &dir, changed, "", 0);
2835 }
2836
2837 static void free_diffstat_info(struct diffstat_t *diffstat)
2838 {
2839 int i;
2840 for (i = 0; i < diffstat->nr; i++) {
2841 struct diffstat_file *f = diffstat->files[i];
2842 if (f->name != f->print_name)
2843 free(f->print_name);
2844 free(f->name);
2845 free(f->from_name);
2846 free(f);
2847 }
2848 free(diffstat->files);
2849 }
2850
2851 struct checkdiff_t {
2852 const char *filename;
2853 int lineno;
2854 int conflict_marker_size;
2855 struct diff_options *o;
2856 unsigned ws_rule;
2857 unsigned status;
2858 };
2859
2860 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2861 {
2862 char firstchar;
2863 int cnt;
2864
2865 if (len < marker_size + 1)
2866 return 0;
2867 firstchar = line[0];
2868 switch (firstchar) {
2869 case '=': case '>': case '<': case '|':
2870 break;
2871 default:
2872 return 0;
2873 }
2874 for (cnt = 1; cnt < marker_size; cnt++)
2875 if (line[cnt] != firstchar)
2876 return 0;
2877 /* line[1] thru line[marker_size-1] are same as firstchar */
2878 if (len < marker_size + 1 || !isspace(line[marker_size]))
2879 return 0;
2880 return 1;
2881 }
2882
2883 static void checkdiff_consume(void *priv, char *line, unsigned long len)
2884 {
2885 struct checkdiff_t *data = priv;
2886 int marker_size = data->conflict_marker_size;
2887 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2888 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2889 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2890 char *err;
2891 const char *line_prefix;
2892
2893 assert(data->o);
2894 line_prefix = diff_line_prefix(data->o);
2895
2896 if (line[0] == '+') {
2897 unsigned bad;
2898 data->lineno++;
2899 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2900 data->status |= 1;
2901 fprintf(data->o->file,
2902 "%s%s:%d: leftover conflict marker\n",
2903 line_prefix, data->filename, data->lineno);
2904 }
2905 bad = ws_check(line + 1, len - 1, data->ws_rule);
2906 if (!bad)
2907 return;
2908 data->status |= bad;
2909 err = whitespace_error_string(bad);
2910 fprintf(data->o->file, "%s%s:%d: %s.\n",
2911 line_prefix, data->filename, data->lineno, err);
2912 free(err);
2913 emit_line(data->o, set, reset, line, 1);
2914 ws_check_emit(line + 1, len - 1, data->ws_rule,
2915 data->o->file, set, reset, ws);
2916 } else if (line[0] == ' ') {
2917 data->lineno++;
2918 } else if (line[0] == '@') {
2919 char *plus = strchr(line, '+');
2920 if (plus)
2921 data->lineno = strtol(plus, NULL, 10) - 1;
2922 else
2923 die("invalid diff");
2924 }
2925 }
2926
2927 static unsigned char *deflate_it(char *data,
2928 unsigned long size,
2929 unsigned long *result_size)
2930 {
2931 int bound;
2932 unsigned char *deflated;
2933 git_zstream stream;
2934
2935 git_deflate_init(&stream, zlib_compression_level);
2936 bound = git_deflate_bound(&stream, size);
2937 deflated = xmalloc(bound);
2938 stream.next_out = deflated;
2939 stream.avail_out = bound;
2940
2941 stream.next_in = (unsigned char *)data;
2942 stream.avail_in = size;
2943 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2944 ; /* nothing */
2945 git_deflate_end(&stream);
2946 *result_size = stream.total_out;
2947 return deflated;
2948 }
2949
2950 static void emit_binary_diff_body(struct diff_options *o,
2951 mmfile_t *one, mmfile_t *two)
2952 {
2953 void *cp;
2954 void *delta;
2955 void *deflated;
2956 void *data;
2957 unsigned long orig_size;
2958 unsigned long delta_size;
2959 unsigned long deflate_size;
2960 unsigned long data_size;
2961
2962 /* We could do deflated delta, or we could do just deflated two,
2963 * whichever is smaller.
2964 */
2965 delta = NULL;
2966 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2967 if (one->size && two->size) {
2968 delta = diff_delta(one->ptr, one->size,
2969 two->ptr, two->size,
2970 &delta_size, deflate_size);
2971 if (delta) {
2972 void *to_free = delta;
2973 orig_size = delta_size;
2974 delta = deflate_it(delta, delta_size, &delta_size);
2975 free(to_free);
2976 }
2977 }
2978
2979 if (delta && delta_size < deflate_size) {
2980 char *s = xstrfmt("%lu", orig_size);
2981 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA,
2982 s, strlen(s), 0);
2983 free(s);
2984 free(deflated);
2985 data = delta;
2986 data_size = delta_size;
2987 } else {
2988 char *s = xstrfmt("%lu", two->size);
2989 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL,
2990 s, strlen(s), 0);
2991 free(s);
2992 free(delta);
2993 data = deflated;
2994 data_size = deflate_size;
2995 }
2996
2997 /* emit data encoded in base85 */
2998 cp = data;
2999 while (data_size) {
3000 int len;
3001 int bytes = (52 < data_size) ? 52 : data_size;
3002 char line[71];
3003 data_size -= bytes;
3004 if (bytes <= 26)
3005 line[0] = bytes + 'A' - 1;
3006 else
3007 line[0] = bytes - 26 + 'a' - 1;
3008 encode_85(line + 1, cp, bytes);
3009 cp = (char *) cp + bytes;
3010
3011 len = strlen(line);
3012 line[len++] = '\n';
3013 line[len] = '\0';
3014
3015 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_BODY,
3016 line, len, 0);
3017 }
3018 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_FOOTER, NULL, 0, 0);
3019 free(data);
3020 }
3021
3022 static void emit_binary_diff(struct diff_options *o,
3023 mmfile_t *one, mmfile_t *two)
3024 {
3025 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER, NULL, 0, 0);
3026 emit_binary_diff_body(o, one, two);
3027 emit_binary_diff_body(o, two, one);
3028 }
3029
3030 int diff_filespec_is_binary(struct diff_filespec *one)
3031 {
3032 if (one->is_binary == -1) {
3033 diff_filespec_load_driver(one);
3034 if (one->driver->binary != -1)
3035 one->is_binary = one->driver->binary;
3036 else {
3037 if (!one->data && DIFF_FILE_VALID(one))
3038 diff_populate_filespec(one, CHECK_BINARY);
3039 if (one->is_binary == -1 && one->data)
3040 one->is_binary = buffer_is_binary(one->data,
3041 one->size);
3042 if (one->is_binary == -1)
3043 one->is_binary = 0;
3044 }
3045 }
3046 return one->is_binary;
3047 }
3048
3049 static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
3050 {
3051 diff_filespec_load_driver(one);
3052 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
3053 }
3054
3055 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
3056 {
3057 if (!options->a_prefix)
3058 options->a_prefix = a;
3059 if (!options->b_prefix)
3060 options->b_prefix = b;
3061 }
3062
3063 struct userdiff_driver *get_textconv(struct diff_filespec *one)
3064 {
3065 if (!DIFF_FILE_VALID(one))
3066 return NULL;
3067
3068 diff_filespec_load_driver(one);
3069 return userdiff_get_textconv(one->driver);
3070 }
3071
3072 static void builtin_diff(const char *name_a,
3073 const char *name_b,
3074 struct diff_filespec *one,
3075 struct diff_filespec *two,
3076 const char *xfrm_msg,
3077 int must_show_header,
3078 struct diff_options *o,
3079 int complete_rewrite)
3080 {
3081 mmfile_t mf1, mf2;
3082 const char *lbl[2];
3083 char *a_one, *b_two;
3084 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
3085 const char *reset = diff_get_color_opt(o, DIFF_RESET);
3086 const char *a_prefix, *b_prefix;
3087 struct userdiff_driver *textconv_one = NULL;
3088 struct userdiff_driver *textconv_two = NULL;
3089 struct strbuf header = STRBUF_INIT;
3090 const char *line_prefix = diff_line_prefix(o);
3091
3092 diff_set_mnemonic_prefix(o, "a/", "b/");
3093 if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
3094 a_prefix = o->b_prefix;
3095 b_prefix = o->a_prefix;
3096 } else {
3097 a_prefix = o->a_prefix;
3098 b_prefix = o->b_prefix;
3099 }
3100
3101 if (o->submodule_format == DIFF_SUBMODULE_LOG &&
3102 (!one->mode || S_ISGITLINK(one->mode)) &&
3103 (!two->mode || S_ISGITLINK(two->mode))) {
3104 show_submodule_summary(o, one->path ? one->path : two->path,
3105 &one->oid, &two->oid,
3106 two->dirty_submodule);
3107 return;
3108 } else if (o->submodule_format == DIFF_SUBMODULE_INLINE_DIFF &&
3109 (!one->mode || S_ISGITLINK(one->mode)) &&
3110 (!two->mode || S_ISGITLINK(two->mode))) {
3111 show_submodule_inline_diff(o, one->path ? one->path : two->path,
3112 &one->oid, &two->oid,
3113 two->dirty_submodule);
3114 return;
3115 }
3116
3117 if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
3118 textconv_one = get_textconv(one);
3119 textconv_two = get_textconv(two);
3120 }
3121
3122 /* Never use a non-valid filename anywhere if at all possible */
3123 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
3124 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
3125
3126 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
3127 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
3128 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
3129 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
3130 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
3131 if (lbl[0][0] == '/') {
3132 /* /dev/null */
3133 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
3134 if (xfrm_msg)
3135 strbuf_addstr(&header, xfrm_msg);
3136 must_show_header = 1;
3137 }
3138 else if (lbl[1][0] == '/') {
3139 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
3140 if (xfrm_msg)
3141 strbuf_addstr(&header, xfrm_msg);
3142 must_show_header = 1;
3143 }
3144 else {
3145 if (one->mode != two->mode) {
3146 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
3147 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
3148 must_show_header = 1;
3149 }
3150 if (xfrm_msg)
3151 strbuf_addstr(&header, xfrm_msg);
3152
3153 /*
3154 * we do not run diff between different kind
3155 * of objects.
3156 */
3157 if ((one->mode ^ two->mode) & S_IFMT)
3158 goto free_ab_and_return;
3159 if (complete_rewrite &&
3160 (textconv_one || !diff_filespec_is_binary(one)) &&
3161 (textconv_two || !diff_filespec_is_binary(two))) {
3162 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3163 header.buf, header.len, 0);
3164 strbuf_reset(&header);
3165 emit_rewrite_diff(name_a, name_b, one, two,
3166 textconv_one, textconv_two, o);
3167 o->found_changes = 1;
3168 goto free_ab_and_return;
3169 }
3170 }
3171
3172 if (o->irreversible_delete && lbl[1][0] == '/') {
3173 emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf,
3174 header.len, 0);
3175 strbuf_reset(&header);
3176 goto free_ab_and_return;
3177 } else if (!DIFF_OPT_TST(o, TEXT) &&
3178 ( (!textconv_one && diff_filespec_is_binary(one)) ||
3179 (!textconv_two && diff_filespec_is_binary(two)) )) {
3180 struct strbuf sb = STRBUF_INIT;
3181 if (!one->data && !two->data &&
3182 S_ISREG(one->mode) && S_ISREG(two->mode) &&
3183 !DIFF_OPT_TST(o, BINARY)) {
3184 if (!oidcmp(&one->oid, &two->oid)) {
3185 if (must_show_header)
3186 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3187 header.buf, header.len,
3188 0);
3189 goto free_ab_and_return;
3190 }
3191 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3192 header.buf, header.len, 0);
3193 strbuf_addf(&sb, "%sBinary files %s and %s differ\n",
3194 diff_line_prefix(o), lbl[0], lbl[1]);
3195 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_FILES,
3196 sb.buf, sb.len, 0);
3197 strbuf_release(&sb);
3198 goto free_ab_and_return;
3199 }
3200 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
3201 die("unable to read files to diff");
3202 /* Quite common confusing case */
3203 if (mf1.size == mf2.size &&
3204 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
3205 if (must_show_header)
3206 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3207 header.buf, header.len, 0);
3208 goto free_ab_and_return;
3209 }
3210 emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf, header.len, 0);
3211 strbuf_reset(&header);
3212 if (DIFF_OPT_TST(o, BINARY))
3213 emit_binary_diff(o, &mf1, &mf2);
3214 else {
3215 strbuf_addf(&sb, "%sBinary files %s and %s differ\n",
3216 diff_line_prefix(o), lbl[0], lbl[1]);
3217 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_FILES,
3218 sb.buf, sb.len, 0);
3219 strbuf_release(&sb);
3220 }
3221 o->found_changes = 1;
3222 } else {
3223 /* Crazy xdl interfaces.. */
3224 const char *diffopts = getenv("GIT_DIFF_OPTS");
3225 const char *v;
3226 xpparam_t xpp;
3227 xdemitconf_t xecfg;
3228 struct emit_callback ecbdata;
3229 const struct userdiff_funcname *pe;
3230
3231 if (must_show_header) {
3232 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3233 header.buf, header.len, 0);
3234 strbuf_reset(&header);
3235 }
3236
3237 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
3238 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
3239
3240 pe = diff_funcname_pattern(one);
3241 if (!pe)
3242 pe = diff_funcname_pattern(two);
3243
3244 memset(&xpp, 0, sizeof(xpp));
3245 memset(&xecfg, 0, sizeof(xecfg));
3246 memset(&ecbdata, 0, sizeof(ecbdata));
3247 ecbdata.label_path = lbl;
3248 ecbdata.color_diff = want_color(o->use_color);
3249 ecbdata.ws_rule = whitespace_rule(name_b);
3250 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
3251 check_blank_at_eof(&mf1, &mf2, &ecbdata);
3252 ecbdata.opt = o;
3253 ecbdata.header = header.len ? &header : NULL;
3254 xpp.flags = o->xdl_opts;
3255 xecfg.ctxlen = o->context;
3256 xecfg.interhunkctxlen = o->interhunkcontext;
3257 xecfg.flags = XDL_EMIT_FUNCNAMES;
3258 if (DIFF_OPT_TST(o, FUNCCONTEXT))
3259 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
3260 if (pe)
3261 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
3262 if (!diffopts)
3263 ;
3264 else if (skip_prefix(diffopts, "--unified=", &v))
3265 xecfg.ctxlen = strtoul(v, NULL, 10);
3266 else if (skip_prefix(diffopts, "-u", &v))
3267 xecfg.ctxlen = strtoul(v, NULL, 10);
3268 if (o->word_diff)
3269 init_diff_words_data(&ecbdata, o, one, two);
3270 if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
3271 &xpp, &xecfg))
3272 die("unable to generate diff for %s", one->path);
3273 if (o->word_diff)
3274 free_diff_words_data(&ecbdata);
3275 if (textconv_one)
3276 free(mf1.ptr);
3277 if (textconv_two)
3278 free(mf2.ptr);
3279 xdiff_clear_find_func(&xecfg);
3280 }
3281
3282 free_ab_and_return:
3283 strbuf_release(&header);
3284 diff_free_filespec_data(one);
3285 diff_free_filespec_data(two);
3286 free(a_one);
3287 free(b_two);
3288 return;
3289 }
3290
3291 static void builtin_diffstat(const char *name_a, const char *name_b,
3292 struct diff_filespec *one,
3293 struct diff_filespec *two,
3294 struct diffstat_t *diffstat,
3295 struct diff_options *o,
3296 struct diff_filepair *p)
3297 {
3298 mmfile_t mf1, mf2;
3299 struct diffstat_file *data;
3300 int same_contents;
3301 int complete_rewrite = 0;
3302
3303 if (!DIFF_PAIR_UNMERGED(p)) {
3304 if (p->status == DIFF_STATUS_MODIFIED && p->score)
3305 complete_rewrite = 1;
3306 }
3307
3308 data = diffstat_add(diffstat, name_a, name_b);
3309 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
3310
3311 if (!one || !two) {
3312 data->is_unmerged = 1;
3313 return;
3314 }
3315
3316 same_contents = !oidcmp(&one->oid, &two->oid);
3317
3318 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
3319 data->is_binary = 1;
3320 if (same_contents) {
3321 data->added = 0;
3322 data->deleted = 0;
3323 } else {
3324 data->added = diff_filespec_size(two);
3325 data->deleted = diff_filespec_size(one);
3326 }
3327 }
3328
3329 else if (complete_rewrite) {
3330 diff_populate_filespec(one, 0);
3331 diff_populate_filespec(two, 0);
3332 data->deleted = count_lines(one->data, one->size);
3333 data->added = count_lines(two->data, two->size);
3334 }
3335
3336 else if (!same_contents) {
3337 /* Crazy xdl interfaces.. */
3338 xpparam_t xpp;
3339 xdemitconf_t xecfg;
3340
3341 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
3342 die("unable to read files to diff");
3343
3344 memset(&xpp, 0, sizeof(xpp));
3345 memset(&xecfg, 0, sizeof(xecfg));
3346 xpp.flags = o->xdl_opts;
3347 xecfg.ctxlen = o->context;
3348 xecfg.interhunkctxlen = o->interhunkcontext;
3349 if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
3350 &xpp, &xecfg))
3351 die("unable to generate diffstat for %s", one->path);
3352 }
3353
3354 diff_free_filespec_data(one);
3355 diff_free_filespec_data(two);
3356 }
3357
3358 static void builtin_checkdiff(const char *name_a, const char *name_b,
3359 const char *attr_path,
3360 struct diff_filespec *one,
3361 struct diff_filespec *two,
3362 struct diff_options *o)
3363 {
3364 mmfile_t mf1, mf2;
3365 struct checkdiff_t data;
3366
3367 if (!two)
3368 return;
3369
3370 memset(&data, 0, sizeof(data));
3371 data.filename = name_b ? name_b : name_a;
3372 data.lineno = 0;
3373 data.o = o;
3374 data.ws_rule = whitespace_rule(attr_path);
3375 data.conflict_marker_size = ll_merge_marker_size(attr_path);
3376
3377 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
3378 die("unable to read files to diff");
3379
3380 /*
3381 * All the other codepaths check both sides, but not checking
3382 * the "old" side here is deliberate. We are checking the newly
3383 * introduced changes, and as long as the "new" side is text, we
3384 * can and should check what it introduces.
3385 */
3386 if (diff_filespec_is_binary(two))
3387 goto free_and_return;
3388 else {
3389 /* Crazy xdl interfaces.. */
3390 xpparam_t xpp;
3391 xdemitconf_t xecfg;
3392
3393 memset(&xpp, 0, sizeof(xpp));
3394 memset(&xecfg, 0, sizeof(xecfg));
3395 xecfg.ctxlen = 1; /* at least one context line */
3396 xpp.flags = 0;
3397 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
3398 &xpp, &xecfg))
3399 die("unable to generate checkdiff for %s", one->path);
3400
3401 if (data.ws_rule & WS_BLANK_AT_EOF) {
3402 struct emit_callback ecbdata;
3403 int blank_at_eof;
3404
3405 ecbdata.ws_rule = data.ws_rule;
3406 check_blank_at_eof(&mf1, &mf2, &ecbdata);
3407 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
3408
3409 if (blank_at_eof) {
3410 static char *err;
3411 if (!err)
3412 err = whitespace_error_string(WS_BLANK_AT_EOF);
3413 fprintf(o->file, "%s:%d: %s.\n",
3414 data.filename, blank_at_eof, err);
3415 data.status = 1; /* report errors */
3416 }
3417 }
3418 }
3419 free_and_return:
3420 diff_free_filespec_data(one);
3421 diff_free_filespec_data(two);
3422 if (data.status)
3423 DIFF_OPT_SET(o, CHECK_FAILED);
3424 }
3425
3426 struct diff_filespec *alloc_filespec(const char *path)
3427 {
3428 struct diff_filespec *spec;
3429
3430 FLEXPTR_ALLOC_STR(spec, path, path);
3431 spec->count = 1;
3432 spec->is_binary = -1;
3433 return spec;
3434 }
3435
3436 void free_filespec(struct diff_filespec *spec)
3437 {
3438 if (!--spec->count) {
3439 diff_free_filespec_data(spec);
3440 free(spec);
3441 }
3442 }
3443
3444 void fill_filespec(struct diff_filespec *spec, const struct object_id *oid,
3445 int oid_valid, unsigned short mode)
3446 {
3447 if (mode) {
3448 spec->mode = canon_mode(mode);
3449 oidcpy(&spec->oid, oid);
3450 spec->oid_valid = oid_valid;
3451 }
3452 }
3453
3454 /*
3455 * Given a name and sha1 pair, if the index tells us the file in
3456 * the work tree has that object contents, return true, so that
3457 * prepare_temp_file() does not have to inflate and extract.
3458 */
3459 static int reuse_worktree_file(const char *name, const struct object_id *oid, int want_file)
3460 {
3461 const struct cache_entry *ce;
3462 struct stat st;
3463 int pos, len;
3464
3465 /*
3466 * We do not read the cache ourselves here, because the
3467 * benchmark with my previous version that always reads cache
3468 * shows that it makes things worse for diff-tree comparing
3469 * two linux-2.6 kernel trees in an already checked out work
3470 * tree. This is because most diff-tree comparisons deal with
3471 * only a small number of files, while reading the cache is
3472 * expensive for a large project, and its cost outweighs the
3473 * savings we get by not inflating the object to a temporary
3474 * file. Practically, this code only helps when we are used
3475 * by diff-cache --cached, which does read the cache before
3476 * calling us.
3477 */
3478 if (!active_cache)
3479 return 0;
3480
3481 /* We want to avoid the working directory if our caller
3482 * doesn't need the data in a normal file, this system
3483 * is rather slow with its stat/open/mmap/close syscalls,
3484 * and the object is contained in a pack file. The pack
3485 * is probably already open and will be faster to obtain
3486 * the data through than the working directory. Loose
3487 * objects however would tend to be slower as they need
3488 * to be individually opened and inflated.
3489 */
3490 if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(oid->hash))
3491 return 0;
3492
3493 /*
3494 * Similarly, if we'd have to convert the file contents anyway, that
3495 * makes the optimization not worthwhile.
3496 */
3497 if (!want_file && would_convert_to_git(&the_index, name))
3498 return 0;
3499
3500 len = strlen(name);
3501 pos = cache_name_pos(name, len);
3502 if (pos < 0)
3503 return 0;
3504 ce = active_cache[pos];
3505
3506 /*
3507 * This is not the sha1 we are looking for, or
3508 * unreusable because it is not a regular file.
3509 */
3510 if (oidcmp(oid, &ce->oid) || !S_ISREG(ce->ce_mode))
3511 return 0;
3512
3513 /*
3514 * If ce is marked as "assume unchanged", there is no
3515 * guarantee that work tree matches what we are looking for.
3516 */
3517 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
3518 return 0;
3519
3520 /*
3521 * If ce matches the file in the work tree, we can reuse it.
3522 */
3523 if (ce_uptodate(ce) ||
3524 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
3525 return 1;
3526
3527 return 0;
3528 }
3529
3530 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
3531 {
3532 struct strbuf buf = STRBUF_INIT;
3533 char *dirty = "";
3534
3535 /* Are we looking at the work tree? */
3536 if (s->dirty_submodule)
3537 dirty = "-dirty";
3538
3539 strbuf_addf(&buf, "Subproject commit %s%s\n",
3540 oid_to_hex(&s->oid), dirty);
3541 s->size = buf.len;
3542 if (size_only) {
3543 s->data = NULL;
3544 strbuf_release(&buf);
3545 } else {
3546 s->data = strbuf_detach(&buf, NULL);
3547 s->should_free = 1;
3548 }
3549 return 0;
3550 }
3551
3552 /*
3553 * While doing rename detection and pickaxe operation, we may need to
3554 * grab the data for the blob (or file) for our own in-core comparison.
3555 * diff_filespec has data and size fields for this purpose.
3556 */
3557 int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
3558 {
3559 int size_only = flags & CHECK_SIZE_ONLY;
3560 int err = 0;
3561 /*
3562 * demote FAIL to WARN to allow inspecting the situation
3563 * instead of refusing.
3564 */
3565 enum safe_crlf crlf_warn = (safe_crlf == SAFE_CRLF_FAIL
3566 ? SAFE_CRLF_WARN
3567 : safe_crlf);
3568
3569 if (!DIFF_FILE_VALID(s))
3570 die("internal error: asking to populate invalid file.");
3571 if (S_ISDIR(s->mode))
3572 return -1;
3573
3574 if (s->data)
3575 return 0;
3576
3577 if (size_only && 0 < s->size)
3578 return 0;
3579
3580 if (S_ISGITLINK(s->mode))
3581 return diff_populate_gitlink(s, size_only);
3582
3583 if (!s->oid_valid ||
3584 reuse_worktree_file(s->path, &s->oid, 0)) {
3585 struct strbuf buf = STRBUF_INIT;
3586 struct stat st;
3587 int fd;
3588
3589 if (lstat(s->path, &st) < 0) {
3590 if (errno == ENOENT) {
3591 err_empty:
3592 err = -1;
3593 empty:
3594 s->data = (char *)"";
3595 s->size = 0;
3596 return err;
3597 }
3598 }
3599 s->size = xsize_t(st.st_size);
3600 if (!s->size)
3601 goto empty;
3602 if (S_ISLNK(st.st_mode)) {
3603 struct strbuf sb = STRBUF_INIT;
3604
3605 if (strbuf_readlink(&sb, s->path, s->size))
3606 goto err_empty;
3607 s->size = sb.len;
3608 s->data = strbuf_detach(&sb, NULL);
3609 s->should_free = 1;
3610 return 0;
3611 }
3612
3613 /*
3614 * Even if the caller would be happy with getting
3615 * only the size, we cannot return early at this
3616 * point if the path requires us to run the content
3617 * conversion.
3618 */
3619 if (size_only && !would_convert_to_git(&the_index, s->path))
3620 return 0;
3621
3622 /*
3623 * Note: this check uses xsize_t(st.st_size) that may
3624 * not be the true size of the blob after it goes
3625 * through convert_to_git(). This may not strictly be
3626 * correct, but the whole point of big_file_threshold
3627 * and is_binary check being that we want to avoid
3628 * opening the file and inspecting the contents, this
3629 * is probably fine.
3630 */
3631 if ((flags & CHECK_BINARY) &&
3632 s->size > big_file_threshold && s->is_binary == -1) {
3633 s->is_binary = 1;
3634 return 0;
3635 }
3636 fd = open(s->path, O_RDONLY);
3637 if (fd < 0)
3638 goto err_empty;
3639 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
3640 close(fd);
3641 s->should_munmap = 1;
3642
3643 /*
3644 * Convert from working tree format to canonical git format
3645 */
3646 if (convert_to_git(&the_index, s->path, s->data, s->size, &buf, crlf_warn)) {
3647 size_t size = 0;
3648 munmap(s->data, s->size);
3649 s->should_munmap = 0;
3650 s->data = strbuf_detach(&buf, &size);
3651 s->size = size;
3652 s->should_free = 1;
3653 }
3654 }
3655 else {
3656 enum object_type type;
3657 if (size_only || (flags & CHECK_BINARY)) {
3658 type = sha1_object_info(s->oid.hash, &s->size);
3659 if (type < 0)
3660 die("unable to read %s",
3661 oid_to_hex(&s->oid));
3662 if (size_only)
3663 return 0;
3664 if (s->size > big_file_threshold && s->is_binary == -1) {
3665 s->is_binary = 1;
3666 return 0;
3667 }
3668 }
3669 s->data = read_sha1_file(s->oid.hash, &type, &s->size);
3670 if (!s->data)
3671 die("unable to read %s", oid_to_hex(&s->oid));
3672 s->should_free = 1;
3673 }
3674 return 0;
3675 }
3676
3677 void diff_free_filespec_blob(struct diff_filespec *s)
3678 {
3679 if (s->should_free)
3680 free(s->data);
3681 else if (s->should_munmap)
3682 munmap(s->data, s->size);
3683
3684 if (s->should_free || s->should_munmap) {
3685 s->should_free = s->should_munmap = 0;
3686 s->data = NULL;
3687 }
3688 }
3689
3690 void diff_free_filespec_data(struct diff_filespec *s)
3691 {
3692 diff_free_filespec_blob(s);
3693 FREE_AND_NULL(s->cnt_data);
3694 }
3695
3696 static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
3697 void *blob,
3698 unsigned long size,
3699 const struct object_id *oid,
3700 int mode)
3701 {
3702 int fd;
3703 struct strbuf buf = STRBUF_INIT;
3704 struct strbuf template = STRBUF_INIT;
3705 char *path_dup = xstrdup(path);
3706 const char *base = basename(path_dup);
3707
3708 /* Generate "XXXXXX_basename.ext" */
3709 strbuf_addstr(&template, "XXXXXX_");
3710 strbuf_addstr(&template, base);
3711
3712 fd = mks_tempfile_ts(&temp->tempfile, template.buf, strlen(base) + 1);
3713 if (fd < 0)
3714 die_errno("unable to create temp-file");
3715 if (convert_to_working_tree(path,
3716 (const char *)blob, (size_t)size, &buf)) {
3717 blob = buf.buf;
3718 size = buf.len;
3719 }
3720 if (write_in_full(fd, blob, size) != size)
3721 die_errno("unable to write temp-file");
3722 close_tempfile(&temp->tempfile);
3723 temp->name = get_tempfile_path(&temp->tempfile);
3724 oid_to_hex_r(temp->hex, oid);
3725 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", mode);
3726 strbuf_release(&buf);
3727 strbuf_release(&template);
3728 free(path_dup);
3729 }
3730
3731 static struct diff_tempfile *prepare_temp_file(const char *name,
3732 struct diff_filespec *one)
3733 {
3734 struct diff_tempfile *temp = claim_diff_tempfile();