Update messages in preparation for i18n
[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 #include "packfile.h"
25 #include "help.h"
26
27 #ifdef NO_FAST_WORKING_DIRECTORY
28 #define FAST_WORKING_DIRECTORY 0
29 #else
30 #define FAST_WORKING_DIRECTORY 1
31 #endif
32
33 static int diff_detect_rename_default;
34 static int diff_indent_heuristic = 1;
35 static int diff_rename_limit_default = 400;
36 static int diff_suppress_blank_empty;
37 static int diff_use_color_default = -1;
38 static int diff_color_moved_default;
39 static int diff_context_default = 3;
40 static int diff_interhunk_context_default;
41 static const char *diff_word_regex_cfg;
42 static const char *external_diff_cmd_cfg;
43 static const char *diff_order_file_cfg;
44 int diff_auto_refresh_index = 1;
45 static int diff_mnemonic_prefix;
46 static int diff_no_prefix;
47 static int diff_stat_graph_width;
48 static int diff_dirstat_permille_default = 30;
49 static struct diff_options default_diff_options;
50 static long diff_algorithm;
51 static unsigned ws_error_highlight_default = WSEH_NEW;
52
53 static char diff_colors[][COLOR_MAXLEN] = {
54 GIT_COLOR_RESET,
55 GIT_COLOR_NORMAL, /* CONTEXT */
56 GIT_COLOR_BOLD, /* METAINFO */
57 GIT_COLOR_CYAN, /* FRAGINFO */
58 GIT_COLOR_RED, /* OLD */
59 GIT_COLOR_GREEN, /* NEW */
60 GIT_COLOR_YELLOW, /* COMMIT */
61 GIT_COLOR_BG_RED, /* WHITESPACE */
62 GIT_COLOR_NORMAL, /* FUNCINFO */
63 GIT_COLOR_BOLD_MAGENTA, /* OLD_MOVED */
64 GIT_COLOR_BOLD_BLUE, /* OLD_MOVED ALTERNATIVE */
65 GIT_COLOR_FAINT, /* OLD_MOVED_DIM */
66 GIT_COLOR_FAINT_ITALIC, /* OLD_MOVED_ALTERNATIVE_DIM */
67 GIT_COLOR_BOLD_CYAN, /* NEW_MOVED */
68 GIT_COLOR_BOLD_YELLOW, /* NEW_MOVED ALTERNATIVE */
69 GIT_COLOR_FAINT, /* NEW_MOVED_DIM */
70 GIT_COLOR_FAINT_ITALIC, /* NEW_MOVED_ALTERNATIVE_DIM */
71 };
72
73 static const char *color_diff_slots[] = {
74 [DIFF_CONTEXT] = "context",
75 [DIFF_METAINFO] = "meta",
76 [DIFF_FRAGINFO] = "frag",
77 [DIFF_FILE_OLD] = "old",
78 [DIFF_FILE_NEW] = "new",
79 [DIFF_COMMIT] = "commit",
80 [DIFF_WHITESPACE] = "whitespace",
81 [DIFF_FUNCINFO] = "func",
82 [DIFF_FILE_OLD_MOVED] = "oldMoved",
83 [DIFF_FILE_OLD_MOVED_ALT] = "oldMovedAlternative",
84 [DIFF_FILE_OLD_MOVED_DIM] = "oldMovedDimmed",
85 [DIFF_FILE_OLD_MOVED_ALT_DIM] = "oldMovedAlternativeDimmed",
86 [DIFF_FILE_NEW_MOVED] = "newMoved",
87 [DIFF_FILE_NEW_MOVED_ALT] = "newMovedAlternative",
88 [DIFF_FILE_NEW_MOVED_DIM] = "newMovedDimmed",
89 [DIFF_FILE_NEW_MOVED_ALT_DIM] = "newMovedAlternativeDimmed",
90 };
91
92 static NORETURN void die_want_option(const char *option_name)
93 {
94 die(_("option '%s' requires a value"), option_name);
95 }
96
97 define_list_config_array_extra(color_diff_slots, {"plain"});
98
99 static int parse_diff_color_slot(const char *var)
100 {
101 if (!strcasecmp(var, "plain"))
102 return DIFF_CONTEXT;
103 return LOOKUP_CONFIG(color_diff_slots, var);
104 }
105
106 static int parse_dirstat_params(struct diff_options *options, const char *params_string,
107 struct strbuf *errmsg)
108 {
109 char *params_copy = xstrdup(params_string);
110 struct string_list params = STRING_LIST_INIT_NODUP;
111 int ret = 0;
112 int i;
113
114 if (*params_copy)
115 string_list_split_in_place(&params, params_copy, ',', -1);
116 for (i = 0; i < params.nr; i++) {
117 const char *p = params.items[i].string;
118 if (!strcmp(p, "changes")) {
119 options->flags.dirstat_by_line = 0;
120 options->flags.dirstat_by_file = 0;
121 } else if (!strcmp(p, "lines")) {
122 options->flags.dirstat_by_line = 1;
123 options->flags.dirstat_by_file = 0;
124 } else if (!strcmp(p, "files")) {
125 options->flags.dirstat_by_line = 0;
126 options->flags.dirstat_by_file = 1;
127 } else if (!strcmp(p, "noncumulative")) {
128 options->flags.dirstat_cumulative = 0;
129 } else if (!strcmp(p, "cumulative")) {
130 options->flags.dirstat_cumulative = 1;
131 } else if (isdigit(*p)) {
132 char *end;
133 int permille = strtoul(p, &end, 10) * 10;
134 if (*end == '.' && isdigit(*++end)) {
135 /* only use first digit */
136 permille += *end - '0';
137 /* .. and ignore any further digits */
138 while (isdigit(*++end))
139 ; /* nothing */
140 }
141 if (!*end)
142 options->dirstat_permille = permille;
143 else {
144 strbuf_addf(errmsg, _(" Failed to parse dirstat cut-off percentage '%s'\n"),
145 p);
146 ret++;
147 }
148 } else {
149 strbuf_addf(errmsg, _(" Unknown dirstat parameter '%s'\n"), p);
150 ret++;
151 }
152
153 }
154 string_list_clear(&params, 0);
155 free(params_copy);
156 return ret;
157 }
158
159 static int parse_submodule_params(struct diff_options *options, const char *value)
160 {
161 if (!strcmp(value, "log"))
162 options->submodule_format = DIFF_SUBMODULE_LOG;
163 else if (!strcmp(value, "short"))
164 options->submodule_format = DIFF_SUBMODULE_SHORT;
165 else if (!strcmp(value, "diff"))
166 options->submodule_format = DIFF_SUBMODULE_INLINE_DIFF;
167 else
168 return -1;
169 return 0;
170 }
171
172 int git_config_rename(const char *var, const char *value)
173 {
174 if (!value)
175 return DIFF_DETECT_RENAME;
176 if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
177 return DIFF_DETECT_COPY;
178 return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
179 }
180
181 long parse_algorithm_value(const char *value)
182 {
183 if (!value)
184 return -1;
185 else if (!strcasecmp(value, "myers") || !strcasecmp(value, "default"))
186 return 0;
187 else if (!strcasecmp(value, "minimal"))
188 return XDF_NEED_MINIMAL;
189 else if (!strcasecmp(value, "patience"))
190 return XDF_PATIENCE_DIFF;
191 else if (!strcasecmp(value, "histogram"))
192 return XDF_HISTOGRAM_DIFF;
193 return -1;
194 }
195
196 static int parse_one_token(const char **arg, const char *token)
197 {
198 const char *rest;
199 if (skip_prefix(*arg, token, &rest) && (!*rest || *rest == ',')) {
200 *arg = rest;
201 return 1;
202 }
203 return 0;
204 }
205
206 static int parse_ws_error_highlight(const char *arg)
207 {
208 const char *orig_arg = arg;
209 unsigned val = 0;
210
211 while (*arg) {
212 if (parse_one_token(&arg, "none"))
213 val = 0;
214 else if (parse_one_token(&arg, "default"))
215 val = WSEH_NEW;
216 else if (parse_one_token(&arg, "all"))
217 val = WSEH_NEW | WSEH_OLD | WSEH_CONTEXT;
218 else if (parse_one_token(&arg, "new"))
219 val |= WSEH_NEW;
220 else if (parse_one_token(&arg, "old"))
221 val |= WSEH_OLD;
222 else if (parse_one_token(&arg, "context"))
223 val |= WSEH_CONTEXT;
224 else {
225 return -1 - (int)(arg - orig_arg);
226 }
227 if (*arg)
228 arg++;
229 }
230 return val;
231 }
232
233 /*
234 * These are to give UI layer defaults.
235 * The core-level commands such as git-diff-files should
236 * never be affected by the setting of diff.renames
237 * the user happens to have in the configuration file.
238 */
239 void init_diff_ui_defaults(void)
240 {
241 diff_detect_rename_default = DIFF_DETECT_RENAME;
242 }
243
244 int git_diff_heuristic_config(const char *var, const char *value, void *cb)
245 {
246 if (!strcmp(var, "diff.indentheuristic"))
247 diff_indent_heuristic = git_config_bool(var, value);
248 return 0;
249 }
250
251 static int parse_color_moved(const char *arg)
252 {
253 switch (git_parse_maybe_bool(arg)) {
254 case 0:
255 return COLOR_MOVED_NO;
256 case 1:
257 return COLOR_MOVED_DEFAULT;
258 default:
259 break;
260 }
261
262 if (!strcmp(arg, "no"))
263 return COLOR_MOVED_NO;
264 else if (!strcmp(arg, "plain"))
265 return COLOR_MOVED_PLAIN;
266 else if (!strcmp(arg, "zebra"))
267 return COLOR_MOVED_ZEBRA;
268 else if (!strcmp(arg, "default"))
269 return COLOR_MOVED_DEFAULT;
270 else if (!strcmp(arg, "dimmed_zebra"))
271 return COLOR_MOVED_ZEBRA_DIM;
272 else
273 return error(_("color moved setting must be one of 'no', 'default', 'zebra', 'dimmed_zebra', 'plain'"));
274 }
275
276 int git_diff_ui_config(const char *var, const char *value, void *cb)
277 {
278 if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
279 diff_use_color_default = git_config_colorbool(var, value);
280 return 0;
281 }
282 if (!strcmp(var, "diff.colormoved")) {
283 int cm = parse_color_moved(value);
284 if (cm < 0)
285 return -1;
286 diff_color_moved_default = cm;
287 return 0;
288 }
289 if (!strcmp(var, "diff.context")) {
290 diff_context_default = git_config_int(var, value);
291 if (diff_context_default < 0)
292 return -1;
293 return 0;
294 }
295 if (!strcmp(var, "diff.interhunkcontext")) {
296 diff_interhunk_context_default = git_config_int(var, value);
297 if (diff_interhunk_context_default < 0)
298 return -1;
299 return 0;
300 }
301 if (!strcmp(var, "diff.renames")) {
302 diff_detect_rename_default = git_config_rename(var, value);
303 return 0;
304 }
305 if (!strcmp(var, "diff.autorefreshindex")) {
306 diff_auto_refresh_index = git_config_bool(var, value);
307 return 0;
308 }
309 if (!strcmp(var, "diff.mnemonicprefix")) {
310 diff_mnemonic_prefix = git_config_bool(var, value);
311 return 0;
312 }
313 if (!strcmp(var, "diff.noprefix")) {
314 diff_no_prefix = git_config_bool(var, value);
315 return 0;
316 }
317 if (!strcmp(var, "diff.statgraphwidth")) {
318 diff_stat_graph_width = git_config_int(var, value);
319 return 0;
320 }
321 if (!strcmp(var, "diff.external"))
322 return git_config_string(&external_diff_cmd_cfg, var, value);
323 if (!strcmp(var, "diff.wordregex"))
324 return git_config_string(&diff_word_regex_cfg, var, value);
325 if (!strcmp(var, "diff.orderfile"))
326 return git_config_pathname(&diff_order_file_cfg, var, value);
327
328 if (!strcmp(var, "diff.ignoresubmodules"))
329 handle_ignore_submodules_arg(&default_diff_options, value);
330
331 if (!strcmp(var, "diff.submodule")) {
332 if (parse_submodule_params(&default_diff_options, value))
333 warning(_("Unknown value for 'diff.submodule' config variable: '%s'"),
334 value);
335 return 0;
336 }
337
338 if (!strcmp(var, "diff.algorithm")) {
339 diff_algorithm = parse_algorithm_value(value);
340 if (diff_algorithm < 0)
341 return -1;
342 return 0;
343 }
344
345 if (!strcmp(var, "diff.wserrorhighlight")) {
346 int val = parse_ws_error_highlight(value);
347 if (val < 0)
348 return -1;
349 ws_error_highlight_default = val;
350 return 0;
351 }
352
353 if (git_color_config(var, value, cb) < 0)
354 return -1;
355
356 return git_diff_basic_config(var, value, cb);
357 }
358
359 int git_diff_basic_config(const char *var, const char *value, void *cb)
360 {
361 const char *name;
362
363 if (!strcmp(var, "diff.renamelimit")) {
364 diff_rename_limit_default = git_config_int(var, value);
365 return 0;
366 }
367
368 if (userdiff_config(var, value) < 0)
369 return -1;
370
371 if (skip_prefix(var, "diff.color.", &name) ||
372 skip_prefix(var, "color.diff.", &name)) {
373 int slot = parse_diff_color_slot(name);
374 if (slot < 0)
375 return 0;
376 if (!value)
377 return config_error_nonbool(var);
378 return color_parse(value, diff_colors[slot]);
379 }
380
381 /* like GNU diff's --suppress-blank-empty option */
382 if (!strcmp(var, "diff.suppressblankempty") ||
383 /* for backwards compatibility */
384 !strcmp(var, "diff.suppress-blank-empty")) {
385 diff_suppress_blank_empty = git_config_bool(var, value);
386 return 0;
387 }
388
389 if (!strcmp(var, "diff.dirstat")) {
390 struct strbuf errmsg = STRBUF_INIT;
391 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
392 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
393 warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
394 errmsg.buf);
395 strbuf_release(&errmsg);
396 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
397 return 0;
398 }
399
400 if (git_diff_heuristic_config(var, value, cb) < 0)
401 return -1;
402
403 return git_default_config(var, value, cb);
404 }
405
406 static char *quote_two(const char *one, const char *two)
407 {
408 int need_one = quote_c_style(one, NULL, NULL, 1);
409 int need_two = quote_c_style(two, NULL, NULL, 1);
410 struct strbuf res = STRBUF_INIT;
411
412 if (need_one + need_two) {
413 strbuf_addch(&res, '"');
414 quote_c_style(one, &res, NULL, 1);
415 quote_c_style(two, &res, NULL, 1);
416 strbuf_addch(&res, '"');
417 } else {
418 strbuf_addstr(&res, one);
419 strbuf_addstr(&res, two);
420 }
421 return strbuf_detach(&res, NULL);
422 }
423
424 static const char *external_diff(void)
425 {
426 static const char *external_diff_cmd = NULL;
427 static int done_preparing = 0;
428
429 if (done_preparing)
430 return external_diff_cmd;
431 external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
432 if (!external_diff_cmd)
433 external_diff_cmd = external_diff_cmd_cfg;
434 done_preparing = 1;
435 return external_diff_cmd;
436 }
437
438 /*
439 * Keep track of files used for diffing. Sometimes such an entry
440 * refers to a temporary file, sometimes to an existing file, and
441 * sometimes to "/dev/null".
442 */
443 static struct diff_tempfile {
444 /*
445 * filename external diff should read from, or NULL if this
446 * entry is currently not in use:
447 */
448 const char *name;
449
450 char hex[GIT_MAX_HEXSZ + 1];
451 char mode[10];
452
453 /*
454 * If this diff_tempfile instance refers to a temporary file,
455 * this tempfile object is used to manage its lifetime.
456 */
457 struct tempfile *tempfile;
458 } diff_temp[2];
459
460 struct emit_callback {
461 int color_diff;
462 unsigned ws_rule;
463 int blank_at_eof_in_preimage;
464 int blank_at_eof_in_postimage;
465 int lno_in_preimage;
466 int lno_in_postimage;
467 const char **label_path;
468 struct diff_words_data *diff_words;
469 struct diff_options *opt;
470 struct strbuf *header;
471 };
472
473 static int count_lines(const char *data, int size)
474 {
475 int count, ch, completely_empty = 1, nl_just_seen = 0;
476 count = 0;
477 while (0 < size--) {
478 ch = *data++;
479 if (ch == '\n') {
480 count++;
481 nl_just_seen = 1;
482 completely_empty = 0;
483 }
484 else {
485 nl_just_seen = 0;
486 completely_empty = 0;
487 }
488 }
489 if (completely_empty)
490 return 0;
491 if (!nl_just_seen)
492 count++; /* no trailing newline */
493 return count;
494 }
495
496 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
497 {
498 if (!DIFF_FILE_VALID(one)) {
499 mf->ptr = (char *)""; /* does not matter */
500 mf->size = 0;
501 return 0;
502 }
503 else if (diff_populate_filespec(one, 0))
504 return -1;
505
506 mf->ptr = one->data;
507 mf->size = one->size;
508 return 0;
509 }
510
511 /* like fill_mmfile, but only for size, so we can avoid retrieving blob */
512 static unsigned long diff_filespec_size(struct diff_filespec *one)
513 {
514 if (!DIFF_FILE_VALID(one))
515 return 0;
516 diff_populate_filespec(one, CHECK_SIZE_ONLY);
517 return one->size;
518 }
519
520 static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
521 {
522 char *ptr = mf->ptr;
523 long size = mf->size;
524 int cnt = 0;
525
526 if (!size)
527 return cnt;
528 ptr += size - 1; /* pointing at the very end */
529 if (*ptr != '\n')
530 ; /* incomplete line */
531 else
532 ptr--; /* skip the last LF */
533 while (mf->ptr < ptr) {
534 char *prev_eol;
535 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
536 if (*prev_eol == '\n')
537 break;
538 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
539 break;
540 cnt++;
541 ptr = prev_eol - 1;
542 }
543 return cnt;
544 }
545
546 static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
547 struct emit_callback *ecbdata)
548 {
549 int l1, l2, at;
550 unsigned ws_rule = ecbdata->ws_rule;
551 l1 = count_trailing_blank(mf1, ws_rule);
552 l2 = count_trailing_blank(mf2, ws_rule);
553 if (l2 <= l1) {
554 ecbdata->blank_at_eof_in_preimage = 0;
555 ecbdata->blank_at_eof_in_postimage = 0;
556 return;
557 }
558 at = count_lines(mf1->ptr, mf1->size);
559 ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
560
561 at = count_lines(mf2->ptr, mf2->size);
562 ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
563 }
564
565 static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
566 int first, const char *line, int len)
567 {
568 int has_trailing_newline, has_trailing_carriage_return;
569 int nofirst;
570 FILE *file = o->file;
571
572 fputs(diff_line_prefix(o), file);
573
574 if (len == 0) {
575 has_trailing_newline = (first == '\n');
576 has_trailing_carriage_return = (!has_trailing_newline &&
577 (first == '\r'));
578 nofirst = has_trailing_newline || has_trailing_carriage_return;
579 } else {
580 has_trailing_newline = (len > 0 && line[len-1] == '\n');
581 if (has_trailing_newline)
582 len--;
583 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
584 if (has_trailing_carriage_return)
585 len--;
586 nofirst = 0;
587 }
588
589 if (len || !nofirst) {
590 fputs(set, file);
591 if (!nofirst)
592 fputc(first, file);
593 fwrite(line, len, 1, file);
594 fputs(reset, file);
595 }
596 if (has_trailing_carriage_return)
597 fputc('\r', file);
598 if (has_trailing_newline)
599 fputc('\n', file);
600 }
601
602 static void emit_line(struct diff_options *o, const char *set, const char *reset,
603 const char *line, int len)
604 {
605 emit_line_0(o, set, reset, line[0], line+1, len-1);
606 }
607
608 enum diff_symbol {
609 DIFF_SYMBOL_BINARY_DIFF_HEADER,
610 DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA,
611 DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL,
612 DIFF_SYMBOL_BINARY_DIFF_BODY,
613 DIFF_SYMBOL_BINARY_DIFF_FOOTER,
614 DIFF_SYMBOL_STATS_SUMMARY_NO_FILES,
615 DIFF_SYMBOL_STATS_SUMMARY_ABBREV,
616 DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES,
617 DIFF_SYMBOL_STATS_LINE,
618 DIFF_SYMBOL_WORD_DIFF,
619 DIFF_SYMBOL_STAT_SEP,
620 DIFF_SYMBOL_SUMMARY,
621 DIFF_SYMBOL_SUBMODULE_ADD,
622 DIFF_SYMBOL_SUBMODULE_DEL,
623 DIFF_SYMBOL_SUBMODULE_UNTRACKED,
624 DIFF_SYMBOL_SUBMODULE_MODIFIED,
625 DIFF_SYMBOL_SUBMODULE_HEADER,
626 DIFF_SYMBOL_SUBMODULE_ERROR,
627 DIFF_SYMBOL_SUBMODULE_PIPETHROUGH,
628 DIFF_SYMBOL_REWRITE_DIFF,
629 DIFF_SYMBOL_BINARY_FILES,
630 DIFF_SYMBOL_HEADER,
631 DIFF_SYMBOL_FILEPAIR_PLUS,
632 DIFF_SYMBOL_FILEPAIR_MINUS,
633 DIFF_SYMBOL_WORDS_PORCELAIN,
634 DIFF_SYMBOL_WORDS,
635 DIFF_SYMBOL_CONTEXT,
636 DIFF_SYMBOL_CONTEXT_INCOMPLETE,
637 DIFF_SYMBOL_PLUS,
638 DIFF_SYMBOL_MINUS,
639 DIFF_SYMBOL_NO_LF_EOF,
640 DIFF_SYMBOL_CONTEXT_FRAGINFO,
641 DIFF_SYMBOL_CONTEXT_MARKER,
642 DIFF_SYMBOL_SEPARATOR
643 };
644 /*
645 * Flags for content lines:
646 * 0..12 are whitespace rules
647 * 13-15 are WSEH_NEW | WSEH_OLD | WSEH_CONTEXT
648 * 16 is marking if the line is blank at EOF
649 */
650 #define DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF (1<<16)
651 #define DIFF_SYMBOL_MOVED_LINE (1<<17)
652 #define DIFF_SYMBOL_MOVED_LINE_ALT (1<<18)
653 #define DIFF_SYMBOL_MOVED_LINE_UNINTERESTING (1<<19)
654 #define DIFF_SYMBOL_CONTENT_WS_MASK (WSEH_NEW | WSEH_OLD | WSEH_CONTEXT | WS_RULE_MASK)
655
656 /*
657 * This struct is used when we need to buffer the output of the diff output.
658 *
659 * NEEDSWORK: Instead of storing a copy of the line, add an offset pointer
660 * into the pre/post image file. This pointer could be a union with the
661 * line pointer. By storing an offset into the file instead of the literal line,
662 * we can decrease the memory footprint for the buffered output. At first we
663 * may want to only have indirection for the content lines, but we could also
664 * enhance the state for emitting prefabricated lines, e.g. the similarity
665 * score line or hunk/file headers would only need to store a number or path
666 * and then the output can be constructed later on depending on state.
667 */
668 struct emitted_diff_symbol {
669 const char *line;
670 int len;
671 int flags;
672 enum diff_symbol s;
673 };
674 #define EMITTED_DIFF_SYMBOL_INIT {NULL}
675
676 struct emitted_diff_symbols {
677 struct emitted_diff_symbol *buf;
678 int nr, alloc;
679 };
680 #define EMITTED_DIFF_SYMBOLS_INIT {NULL, 0, 0}
681
682 static void append_emitted_diff_symbol(struct diff_options *o,
683 struct emitted_diff_symbol *e)
684 {
685 struct emitted_diff_symbol *f;
686
687 ALLOC_GROW(o->emitted_symbols->buf,
688 o->emitted_symbols->nr + 1,
689 o->emitted_symbols->alloc);
690 f = &o->emitted_symbols->buf[o->emitted_symbols->nr++];
691
692 memcpy(f, e, sizeof(struct emitted_diff_symbol));
693 f->line = e->line ? xmemdupz(e->line, e->len) : NULL;
694 }
695
696 struct moved_entry {
697 struct hashmap_entry ent;
698 const struct emitted_diff_symbol *es;
699 struct moved_entry *next_line;
700 };
701
702 static int moved_entry_cmp(const struct diff_options *diffopt,
703 const struct moved_entry *a,
704 const struct moved_entry *b,
705 const void *keydata)
706 {
707 return !xdiff_compare_lines(a->es->line, a->es->len,
708 b->es->line, b->es->len,
709 diffopt->xdl_opts);
710 }
711
712 static struct moved_entry *prepare_entry(struct diff_options *o,
713 int line_no)
714 {
715 struct moved_entry *ret = xmalloc(sizeof(*ret));
716 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[line_no];
717
718 ret->ent.hash = xdiff_hash_string(l->line, l->len, o->xdl_opts);
719 ret->es = l;
720 ret->next_line = NULL;
721
722 return ret;
723 }
724
725 static void add_lines_to_move_detection(struct diff_options *o,
726 struct hashmap *add_lines,
727 struct hashmap *del_lines)
728 {
729 struct moved_entry *prev_line = NULL;
730
731 int n;
732 for (n = 0; n < o->emitted_symbols->nr; n++) {
733 struct hashmap *hm;
734 struct moved_entry *key;
735
736 switch (o->emitted_symbols->buf[n].s) {
737 case DIFF_SYMBOL_PLUS:
738 hm = add_lines;
739 break;
740 case DIFF_SYMBOL_MINUS:
741 hm = del_lines;
742 break;
743 default:
744 prev_line = NULL;
745 continue;
746 }
747
748 key = prepare_entry(o, n);
749 if (prev_line && prev_line->es->s == o->emitted_symbols->buf[n].s)
750 prev_line->next_line = key;
751
752 hashmap_add(hm, key);
753 prev_line = key;
754 }
755 }
756
757 static int shrink_potential_moved_blocks(struct moved_entry **pmb,
758 int pmb_nr)
759 {
760 int lp, rp;
761
762 /* Shrink the set of potential block to the remaining running */
763 for (lp = 0, rp = pmb_nr - 1; lp <= rp;) {
764 while (lp < pmb_nr && pmb[lp])
765 lp++;
766 /* lp points at the first NULL now */
767
768 while (rp > -1 && !pmb[rp])
769 rp--;
770 /* rp points at the last non-NULL */
771
772 if (lp < pmb_nr && rp > -1 && lp < rp) {
773 pmb[lp] = pmb[rp];
774 pmb[rp] = NULL;
775 rp--;
776 lp++;
777 }
778 }
779
780 /* Remember the number of running sets */
781 return rp + 1;
782 }
783
784 /*
785 * If o->color_moved is COLOR_MOVED_PLAIN, this function does nothing.
786 *
787 * Otherwise, if the last block has fewer alphanumeric characters than
788 * COLOR_MOVED_MIN_ALNUM_COUNT, unset DIFF_SYMBOL_MOVED_LINE on all lines in
789 * that block.
790 *
791 * The last block consists of the (n - block_length)'th line up to but not
792 * including the nth line.
793 *
794 * NEEDSWORK: This uses the same heuristic as blame_entry_score() in blame.c.
795 * Think of a way to unify them.
796 */
797 static void adjust_last_block(struct diff_options *o, int n, int block_length)
798 {
799 int i, alnum_count = 0;
800 if (o->color_moved == COLOR_MOVED_PLAIN)
801 return;
802 for (i = 1; i < block_length + 1; i++) {
803 const char *c = o->emitted_symbols->buf[n - i].line;
804 for (; *c; c++) {
805 if (!isalnum(*c))
806 continue;
807 alnum_count++;
808 if (alnum_count >= COLOR_MOVED_MIN_ALNUM_COUNT)
809 return;
810 }
811 }
812 for (i = 1; i < block_length + 1; i++)
813 o->emitted_symbols->buf[n - i].flags &= ~DIFF_SYMBOL_MOVED_LINE;
814 }
815
816 /* Find blocks of moved code, delegate actual coloring decision to helper */
817 static void mark_color_as_moved(struct diff_options *o,
818 struct hashmap *add_lines,
819 struct hashmap *del_lines)
820 {
821 struct moved_entry **pmb = NULL; /* potentially moved blocks */
822 int pmb_nr = 0, pmb_alloc = 0;
823 int n, flipped_block = 1, block_length = 0;
824
825
826 for (n = 0; n < o->emitted_symbols->nr; n++) {
827 struct hashmap *hm = NULL;
828 struct moved_entry *key;
829 struct moved_entry *match = NULL;
830 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[n];
831 int i;
832
833 switch (l->s) {
834 case DIFF_SYMBOL_PLUS:
835 hm = del_lines;
836 key = prepare_entry(o, n);
837 match = hashmap_get(hm, key, o);
838 free(key);
839 break;
840 case DIFF_SYMBOL_MINUS:
841 hm = add_lines;
842 key = prepare_entry(o, n);
843 match = hashmap_get(hm, key, o);
844 free(key);
845 break;
846 default:
847 flipped_block = 1;
848 }
849
850 if (!match) {
851 adjust_last_block(o, n, block_length);
852 pmb_nr = 0;
853 block_length = 0;
854 continue;
855 }
856
857 l->flags |= DIFF_SYMBOL_MOVED_LINE;
858
859 if (o->color_moved == COLOR_MOVED_PLAIN)
860 continue;
861
862 /* Check any potential block runs, advance each or nullify */
863 for (i = 0; i < pmb_nr; i++) {
864 struct moved_entry *p = pmb[i];
865 struct moved_entry *pnext = (p && p->next_line) ?
866 p->next_line : NULL;
867 if (pnext && !hm->cmpfn(o, pnext, match, NULL)) {
868 pmb[i] = p->next_line;
869 } else {
870 pmb[i] = NULL;
871 }
872 }
873
874 pmb_nr = shrink_potential_moved_blocks(pmb, pmb_nr);
875
876 if (pmb_nr == 0) {
877 /*
878 * The current line is the start of a new block.
879 * Setup the set of potential blocks.
880 */
881 for (; match; match = hashmap_get_next(hm, match)) {
882 ALLOC_GROW(pmb, pmb_nr + 1, pmb_alloc);
883 pmb[pmb_nr++] = match;
884 }
885
886 flipped_block = (flipped_block + 1) % 2;
887
888 adjust_last_block(o, n, block_length);
889 block_length = 0;
890 }
891
892 block_length++;
893
894 if (flipped_block)
895 l->flags |= DIFF_SYMBOL_MOVED_LINE_ALT;
896 }
897 adjust_last_block(o, n, block_length);
898
899 free(pmb);
900 }
901
902 #define DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK \
903 (DIFF_SYMBOL_MOVED_LINE | DIFF_SYMBOL_MOVED_LINE_ALT)
904 static void dim_moved_lines(struct diff_options *o)
905 {
906 int n;
907 for (n = 0; n < o->emitted_symbols->nr; n++) {
908 struct emitted_diff_symbol *prev = (n != 0) ?
909 &o->emitted_symbols->buf[n - 1] : NULL;
910 struct emitted_diff_symbol *l = &o->emitted_symbols->buf[n];
911 struct emitted_diff_symbol *next =
912 (n < o->emitted_symbols->nr - 1) ?
913 &o->emitted_symbols->buf[n + 1] : NULL;
914
915 /* Not a plus or minus line? */
916 if (l->s != DIFF_SYMBOL_PLUS && l->s != DIFF_SYMBOL_MINUS)
917 continue;
918
919 /* Not a moved line? */
920 if (!(l->flags & DIFF_SYMBOL_MOVED_LINE))
921 continue;
922
923 /*
924 * If prev or next are not a plus or minus line,
925 * pretend they don't exist
926 */
927 if (prev && prev->s != DIFF_SYMBOL_PLUS &&
928 prev->s != DIFF_SYMBOL_MINUS)
929 prev = NULL;
930 if (next && next->s != DIFF_SYMBOL_PLUS &&
931 next->s != DIFF_SYMBOL_MINUS)
932 next = NULL;
933
934 /* Inside a block? */
935 if ((prev &&
936 (prev->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK) ==
937 (l->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK)) &&
938 (next &&
939 (next->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK) ==
940 (l->flags & DIFF_SYMBOL_MOVED_LINE_ZEBRA_MASK))) {
941 l->flags |= DIFF_SYMBOL_MOVED_LINE_UNINTERESTING;
942 continue;
943 }
944
945 /* Check if we are at an interesting bound: */
946 if (prev && (prev->flags & DIFF_SYMBOL_MOVED_LINE) &&
947 (prev->flags & DIFF_SYMBOL_MOVED_LINE_ALT) !=
948 (l->flags & DIFF_SYMBOL_MOVED_LINE_ALT))
949 continue;
950 if (next && (next->flags & DIFF_SYMBOL_MOVED_LINE) &&
951 (next->flags & DIFF_SYMBOL_MOVED_LINE_ALT) !=
952 (l->flags & DIFF_SYMBOL_MOVED_LINE_ALT))
953 continue;
954
955 /*
956 * The boundary to prev and next are not interesting,
957 * so this line is not interesting as a whole
958 */
959 l->flags |= DIFF_SYMBOL_MOVED_LINE_UNINTERESTING;
960 }
961 }
962
963 static void emit_line_ws_markup(struct diff_options *o,
964 const char *set, const char *reset,
965 const char *line, int len, char sign,
966 unsigned ws_rule, int blank_at_eof)
967 {
968 const char *ws = NULL;
969
970 if (o->ws_error_highlight & ws_rule) {
971 ws = diff_get_color_opt(o, DIFF_WHITESPACE);
972 if (!*ws)
973 ws = NULL;
974 }
975
976 if (!ws)
977 emit_line_0(o, set, reset, sign, line, len);
978 else if (blank_at_eof)
979 /* Blank line at EOF - paint '+' as well */
980 emit_line_0(o, ws, reset, sign, line, len);
981 else {
982 /* Emit just the prefix, then the rest. */
983 emit_line_0(o, set, reset, sign, "", 0);
984 ws_check_emit(line, len, ws_rule,
985 o->file, set, reset, ws);
986 }
987 }
988
989 static void emit_diff_symbol_from_struct(struct diff_options *o,
990 struct emitted_diff_symbol *eds)
991 {
992 static const char *nneof = " No newline at end of file\n";
993 const char *context, *reset, *set, *meta, *fraginfo;
994 struct strbuf sb = STRBUF_INIT;
995
996 enum diff_symbol s = eds->s;
997 const char *line = eds->line;
998 int len = eds->len;
999 unsigned flags = eds->flags;
1000
1001 switch (s) {
1002 case DIFF_SYMBOL_NO_LF_EOF:
1003 context = diff_get_color_opt(o, DIFF_CONTEXT);
1004 reset = diff_get_color_opt(o, DIFF_RESET);
1005 putc('\n', o->file);
1006 emit_line_0(o, context, reset, '\\',
1007 nneof, strlen(nneof));
1008 break;
1009 case DIFF_SYMBOL_SUBMODULE_HEADER:
1010 case DIFF_SYMBOL_SUBMODULE_ERROR:
1011 case DIFF_SYMBOL_SUBMODULE_PIPETHROUGH:
1012 case DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES:
1013 case DIFF_SYMBOL_SUMMARY:
1014 case DIFF_SYMBOL_STATS_LINE:
1015 case DIFF_SYMBOL_BINARY_DIFF_BODY:
1016 case DIFF_SYMBOL_CONTEXT_FRAGINFO:
1017 emit_line(o, "", "", line, len);
1018 break;
1019 case DIFF_SYMBOL_CONTEXT_INCOMPLETE:
1020 case DIFF_SYMBOL_CONTEXT_MARKER:
1021 context = diff_get_color_opt(o, DIFF_CONTEXT);
1022 reset = diff_get_color_opt(o, DIFF_RESET);
1023 emit_line(o, context, reset, line, len);
1024 break;
1025 case DIFF_SYMBOL_SEPARATOR:
1026 fprintf(o->file, "%s%c",
1027 diff_line_prefix(o),
1028 o->line_termination);
1029 break;
1030 case DIFF_SYMBOL_CONTEXT:
1031 set = diff_get_color_opt(o, DIFF_CONTEXT);
1032 reset = diff_get_color_opt(o, DIFF_RESET);
1033 emit_line_ws_markup(o, set, reset, line, len, ' ',
1034 flags & (DIFF_SYMBOL_CONTENT_WS_MASK), 0);
1035 break;
1036 case DIFF_SYMBOL_PLUS:
1037 switch (flags & (DIFF_SYMBOL_MOVED_LINE |
1038 DIFF_SYMBOL_MOVED_LINE_ALT |
1039 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING)) {
1040 case DIFF_SYMBOL_MOVED_LINE |
1041 DIFF_SYMBOL_MOVED_LINE_ALT |
1042 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1043 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_ALT_DIM);
1044 break;
1045 case DIFF_SYMBOL_MOVED_LINE |
1046 DIFF_SYMBOL_MOVED_LINE_ALT:
1047 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_ALT);
1048 break;
1049 case DIFF_SYMBOL_MOVED_LINE |
1050 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1051 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED_DIM);
1052 break;
1053 case DIFF_SYMBOL_MOVED_LINE:
1054 set = diff_get_color_opt(o, DIFF_FILE_NEW_MOVED);
1055 break;
1056 default:
1057 set = diff_get_color_opt(o, DIFF_FILE_NEW);
1058 }
1059 reset = diff_get_color_opt(o, DIFF_RESET);
1060 emit_line_ws_markup(o, set, reset, line, len, '+',
1061 flags & DIFF_SYMBOL_CONTENT_WS_MASK,
1062 flags & DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF);
1063 break;
1064 case DIFF_SYMBOL_MINUS:
1065 switch (flags & (DIFF_SYMBOL_MOVED_LINE |
1066 DIFF_SYMBOL_MOVED_LINE_ALT |
1067 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING)) {
1068 case DIFF_SYMBOL_MOVED_LINE |
1069 DIFF_SYMBOL_MOVED_LINE_ALT |
1070 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1071 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_ALT_DIM);
1072 break;
1073 case DIFF_SYMBOL_MOVED_LINE |
1074 DIFF_SYMBOL_MOVED_LINE_ALT:
1075 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_ALT);
1076 break;
1077 case DIFF_SYMBOL_MOVED_LINE |
1078 DIFF_SYMBOL_MOVED_LINE_UNINTERESTING:
1079 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED_DIM);
1080 break;
1081 case DIFF_SYMBOL_MOVED_LINE:
1082 set = diff_get_color_opt(o, DIFF_FILE_OLD_MOVED);
1083 break;
1084 default:
1085 set = diff_get_color_opt(o, DIFF_FILE_OLD);
1086 }
1087 reset = diff_get_color_opt(o, DIFF_RESET);
1088 emit_line_ws_markup(o, set, reset, line, len, '-',
1089 flags & DIFF_SYMBOL_CONTENT_WS_MASK, 0);
1090 break;
1091 case DIFF_SYMBOL_WORDS_PORCELAIN:
1092 context = diff_get_color_opt(o, DIFF_CONTEXT);
1093 reset = diff_get_color_opt(o, DIFF_RESET);
1094 emit_line(o, context, reset, line, len);
1095 fputs("~\n", o->file);
1096 break;
1097 case DIFF_SYMBOL_WORDS:
1098 context = diff_get_color_opt(o, DIFF_CONTEXT);
1099 reset = diff_get_color_opt(o, DIFF_RESET);
1100 /*
1101 * Skip the prefix character, if any. With
1102 * diff_suppress_blank_empty, there may be
1103 * none.
1104 */
1105 if (line[0] != '\n') {
1106 line++;
1107 len--;
1108 }
1109 emit_line(o, context, reset, line, len);
1110 break;
1111 case DIFF_SYMBOL_FILEPAIR_PLUS:
1112 meta = diff_get_color_opt(o, DIFF_METAINFO);
1113 reset = diff_get_color_opt(o, DIFF_RESET);
1114 fprintf(o->file, "%s%s+++ %s%s%s\n", diff_line_prefix(o), meta,
1115 line, reset,
1116 strchr(line, ' ') ? "\t" : "");
1117 break;
1118 case DIFF_SYMBOL_FILEPAIR_MINUS:
1119 meta = diff_get_color_opt(o, DIFF_METAINFO);
1120 reset = diff_get_color_opt(o, DIFF_RESET);
1121 fprintf(o->file, "%s%s--- %s%s%s\n", diff_line_prefix(o), meta,
1122 line, reset,
1123 strchr(line, ' ') ? "\t" : "");
1124 break;
1125 case DIFF_SYMBOL_BINARY_FILES:
1126 case DIFF_SYMBOL_HEADER:
1127 fprintf(o->file, "%s", line);
1128 break;
1129 case DIFF_SYMBOL_BINARY_DIFF_HEADER:
1130 fprintf(o->file, "%sGIT binary patch\n", diff_line_prefix(o));
1131 break;
1132 case DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA:
1133 fprintf(o->file, "%sdelta %s\n", diff_line_prefix(o), line);
1134 break;
1135 case DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL:
1136 fprintf(o->file, "%sliteral %s\n", diff_line_prefix(o), line);
1137 break;
1138 case DIFF_SYMBOL_BINARY_DIFF_FOOTER:
1139 fputs(diff_line_prefix(o), o->file);
1140 fputc('\n', o->file);
1141 break;
1142 case DIFF_SYMBOL_REWRITE_DIFF:
1143 fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
1144 reset = diff_get_color_opt(o, DIFF_RESET);
1145 emit_line(o, fraginfo, reset, line, len);
1146 break;
1147 case DIFF_SYMBOL_SUBMODULE_ADD:
1148 set = diff_get_color_opt(o, DIFF_FILE_NEW);
1149 reset = diff_get_color_opt(o, DIFF_RESET);
1150 emit_line(o, set, reset, line, len);
1151 break;
1152 case DIFF_SYMBOL_SUBMODULE_DEL:
1153 set = diff_get_color_opt(o, DIFF_FILE_OLD);
1154 reset = diff_get_color_opt(o, DIFF_RESET);
1155 emit_line(o, set, reset, line, len);
1156 break;
1157 case DIFF_SYMBOL_SUBMODULE_UNTRACKED:
1158 fprintf(o->file, "%sSubmodule %s contains untracked content\n",
1159 diff_line_prefix(o), line);
1160 break;
1161 case DIFF_SYMBOL_SUBMODULE_MODIFIED:
1162 fprintf(o->file, "%sSubmodule %s contains modified content\n",
1163 diff_line_prefix(o), line);
1164 break;
1165 case DIFF_SYMBOL_STATS_SUMMARY_NO_FILES:
1166 emit_line(o, "", "", " 0 files changed\n",
1167 strlen(" 0 files changed\n"));
1168 break;
1169 case DIFF_SYMBOL_STATS_SUMMARY_ABBREV:
1170 emit_line(o, "", "", " ...\n", strlen(" ...\n"));
1171 break;
1172 case DIFF_SYMBOL_WORD_DIFF:
1173 fprintf(o->file, "%.*s", len, line);
1174 break;
1175 case DIFF_SYMBOL_STAT_SEP:
1176 fputs(o->stat_sep, o->file);
1177 break;
1178 default:
1179 BUG("unknown diff symbol");
1180 }
1181 strbuf_release(&sb);
1182 }
1183
1184 static void emit_diff_symbol(struct diff_options *o, enum diff_symbol s,
1185 const char *line, int len, unsigned flags)
1186 {
1187 struct emitted_diff_symbol e = {line, len, flags, s};
1188
1189 if (o->emitted_symbols)
1190 append_emitted_diff_symbol(o, &e);
1191 else
1192 emit_diff_symbol_from_struct(o, &e);
1193 }
1194
1195 void diff_emit_submodule_del(struct diff_options *o, const char *line)
1196 {
1197 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_DEL, line, strlen(line), 0);
1198 }
1199
1200 void diff_emit_submodule_add(struct diff_options *o, const char *line)
1201 {
1202 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_ADD, line, strlen(line), 0);
1203 }
1204
1205 void diff_emit_submodule_untracked(struct diff_options *o, const char *path)
1206 {
1207 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_UNTRACKED,
1208 path, strlen(path), 0);
1209 }
1210
1211 void diff_emit_submodule_modified(struct diff_options *o, const char *path)
1212 {
1213 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_MODIFIED,
1214 path, strlen(path), 0);
1215 }
1216
1217 void diff_emit_submodule_header(struct diff_options *o, const char *header)
1218 {
1219 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_HEADER,
1220 header, strlen(header), 0);
1221 }
1222
1223 void diff_emit_submodule_error(struct diff_options *o, const char *err)
1224 {
1225 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_ERROR, err, strlen(err), 0);
1226 }
1227
1228 void diff_emit_submodule_pipethrough(struct diff_options *o,
1229 const char *line, int len)
1230 {
1231 emit_diff_symbol(o, DIFF_SYMBOL_SUBMODULE_PIPETHROUGH, line, len, 0);
1232 }
1233
1234 static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
1235 {
1236 if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
1237 ecbdata->blank_at_eof_in_preimage &&
1238 ecbdata->blank_at_eof_in_postimage &&
1239 ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
1240 ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
1241 return 0;
1242 return ws_blank_line(line, len, ecbdata->ws_rule);
1243 }
1244
1245 static void emit_add_line(const char *reset,
1246 struct emit_callback *ecbdata,
1247 const char *line, int len)
1248 {
1249 unsigned flags = WSEH_NEW | ecbdata->ws_rule;
1250 if (new_blank_line_at_eof(ecbdata, line, len))
1251 flags |= DIFF_SYMBOL_CONTENT_BLANK_LINE_EOF;
1252
1253 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_PLUS, line, len, flags);
1254 }
1255
1256 static void emit_del_line(const char *reset,
1257 struct emit_callback *ecbdata,
1258 const char *line, int len)
1259 {
1260 unsigned flags = WSEH_OLD | ecbdata->ws_rule;
1261 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_MINUS, line, len, flags);
1262 }
1263
1264 static void emit_context_line(const char *reset,
1265 struct emit_callback *ecbdata,
1266 const char *line, int len)
1267 {
1268 unsigned flags = WSEH_CONTEXT | ecbdata->ws_rule;
1269 emit_diff_symbol(ecbdata->opt, DIFF_SYMBOL_CONTEXT, line, len, flags);
1270 }
1271
1272 static void emit_hunk_header(struct emit_callback *ecbdata,
1273 const char *line, int len)
1274 {
1275 const char *context = diff_get_color(ecbdata->color_diff, DIFF_CONTEXT);
1276 const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
1277 const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
1278 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1279 static const char atat[2] = { '@', '@' };
1280 const char *cp, *ep;
1281 struct strbuf msgbuf = STRBUF_INIT;
1282 int org_len = len;
1283 int i = 1;
1284
1285 /*
1286 * As a hunk header must begin with "@@ -<old>, +<new> @@",
1287 * it always is at least 10 bytes long.
1288 */
1289 if (len < 10 ||
1290 memcmp(line, atat, 2) ||
1291 !(ep = memmem(line + 2, len - 2, atat, 2))) {
1292 emit_diff_symbol(ecbdata->opt,
1293 DIFF_SYMBOL_CONTEXT_MARKER, line, len, 0);
1294 return;
1295 }
1296 ep += 2; /* skip over @@ */
1297
1298 /* The hunk header in fraginfo color */
1299 strbuf_addstr(&msgbuf, frag);
1300 strbuf_add(&msgbuf, line, ep - line);
1301 strbuf_addstr(&msgbuf, reset);
1302
1303 /*
1304 * trailing "\r\n"
1305 */
1306 for ( ; i < 3; i++)
1307 if (line[len - i] == '\r' || line[len - i] == '\n')
1308 len--;
1309
1310 /* blank before the func header */
1311 for (cp = ep; ep - line < len; ep++)
1312 if (*ep != ' ' && *ep != '\t')
1313 break;
1314 if (ep != cp) {
1315 strbuf_addstr(&msgbuf, context);
1316 strbuf_add(&msgbuf, cp, ep - cp);
1317 strbuf_addstr(&msgbuf, reset);
1318 }
1319
1320 if (ep < line + len) {
1321 strbuf_addstr(&msgbuf, func);
1322 strbuf_add(&msgbuf, ep, line + len - ep);
1323 strbuf_addstr(&msgbuf, reset);
1324 }
1325
1326 strbuf_add(&msgbuf, line + len, org_len - len);
1327 strbuf_complete_line(&msgbuf);
1328 emit_diff_symbol(ecbdata->opt,
1329 DIFF_SYMBOL_CONTEXT_FRAGINFO, msgbuf.buf, msgbuf.len, 0);
1330 strbuf_release(&msgbuf);
1331 }
1332
1333 static struct diff_tempfile *claim_diff_tempfile(void) {
1334 int i;
1335 for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
1336 if (!diff_temp[i].name)
1337 return diff_temp + i;
1338 BUG("diff is failing to clean up its tempfiles");
1339 }
1340
1341 static void remove_tempfile(void)
1342 {
1343 int i;
1344 for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
1345 if (is_tempfile_active(diff_temp[i].tempfile))
1346 delete_tempfile(&diff_temp[i].tempfile);
1347 diff_temp[i].name = NULL;
1348 }
1349 }
1350
1351 static void add_line_count(struct strbuf *out, int count)
1352 {
1353 switch (count) {
1354 case 0:
1355 strbuf_addstr(out, "0,0");
1356 break;
1357 case 1:
1358 strbuf_addstr(out, "1");
1359 break;
1360 default:
1361 strbuf_addf(out, "1,%d", count);
1362 break;
1363 }
1364 }
1365
1366 static void emit_rewrite_lines(struct emit_callback *ecb,
1367 int prefix, const char *data, int size)
1368 {
1369 const char *endp = NULL;
1370 const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
1371
1372 while (0 < size) {
1373 int len;
1374
1375 endp = memchr(data, '\n', size);
1376 len = endp ? (endp - data + 1) : size;
1377 if (prefix != '+') {
1378 ecb->lno_in_preimage++;
1379 emit_del_line(reset, ecb, data, len);
1380 } else {
1381 ecb->lno_in_postimage++;
1382 emit_add_line(reset, ecb, data, len);
1383 }
1384 size -= len;
1385 data += len;
1386 }
1387 if (!endp)
1388 emit_diff_symbol(ecb->opt, DIFF_SYMBOL_NO_LF_EOF, NULL, 0, 0);
1389 }
1390
1391 static void emit_rewrite_diff(const char *name_a,
1392 const char *name_b,
1393 struct diff_filespec *one,
1394 struct diff_filespec *two,
1395 struct userdiff_driver *textconv_one,
1396 struct userdiff_driver *textconv_two,
1397 struct diff_options *o)
1398 {
1399 int lc_a, lc_b;
1400 static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
1401 const char *a_prefix, *b_prefix;
1402 char *data_one, *data_two;
1403 size_t size_one, size_two;
1404 struct emit_callback ecbdata;
1405 struct strbuf out = STRBUF_INIT;
1406
1407 if (diff_mnemonic_prefix && o->flags.reverse_diff) {
1408 a_prefix = o->b_prefix;
1409 b_prefix = o->a_prefix;
1410 } else {
1411 a_prefix = o->a_prefix;
1412 b_prefix = o->b_prefix;
1413 }
1414
1415 name_a += (*name_a == '/');
1416 name_b += (*name_b == '/');
1417
1418 strbuf_reset(&a_name);
1419 strbuf_reset(&b_name);
1420 quote_two_c_style(&a_name, a_prefix, name_a, 0);
1421 quote_two_c_style(&b_name, b_prefix, name_b, 0);
1422
1423 size_one = fill_textconv(textconv_one, one, &data_one);
1424 size_two = fill_textconv(textconv_two, two, &data_two);
1425
1426 memset(&ecbdata, 0, sizeof(ecbdata));
1427 ecbdata.color_diff = want_color(o->use_color);
1428 ecbdata.ws_rule = whitespace_rule(name_b);
1429 ecbdata.opt = o;
1430 if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
1431 mmfile_t mf1, mf2;
1432 mf1.ptr = (char *)data_one;
1433 mf2.ptr = (char *)data_two;
1434 mf1.size = size_one;
1435 mf2.size = size_two;
1436 check_blank_at_eof(&mf1, &mf2, &ecbdata);
1437 }
1438 ecbdata.lno_in_preimage = 1;
1439 ecbdata.lno_in_postimage = 1;
1440
1441 lc_a = count_lines(data_one, size_one);
1442 lc_b = count_lines(data_two, size_two);
1443
1444 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
1445 a_name.buf, a_name.len, 0);
1446 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
1447 b_name.buf, b_name.len, 0);
1448
1449 strbuf_addstr(&out, "@@ -");
1450 if (!o->irreversible_delete)
1451 add_line_count(&out, lc_a);
1452 else
1453 strbuf_addstr(&out, "?,?");
1454 strbuf_addstr(&out, " +");
1455 add_line_count(&out, lc_b);
1456 strbuf_addstr(&out, " @@\n");
1457 emit_diff_symbol(o, DIFF_SYMBOL_REWRITE_DIFF, out.buf, out.len, 0);
1458 strbuf_release(&out);
1459
1460 if (lc_a && !o->irreversible_delete)
1461 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
1462 if (lc_b)
1463 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
1464 if (textconv_one)
1465 free((char *)data_one);
1466 if (textconv_two)
1467 free((char *)data_two);
1468 }
1469
1470 struct diff_words_buffer {
1471 mmfile_t text;
1472 unsigned long alloc;
1473 struct diff_words_orig {
1474 const char *begin, *end;
1475 } *orig;
1476 int orig_nr, orig_alloc;
1477 };
1478
1479 static void diff_words_append(char *line, unsigned long len,
1480 struct diff_words_buffer *buffer)
1481 {
1482 ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
1483 line++;
1484 len--;
1485 memcpy(buffer->text.ptr + buffer->text.size, line, len);
1486 buffer->text.size += len;
1487 buffer->text.ptr[buffer->text.size] = '\0';
1488 }
1489
1490 struct diff_words_style_elem {
1491 const char *prefix;
1492 const char *suffix;
1493 const char *color; /* NULL; filled in by the setup code if
1494 * color is enabled */
1495 };
1496
1497 struct diff_words_style {
1498 enum diff_words_type type;
1499 struct diff_words_style_elem new_word, old_word, ctx;
1500 const char *newline;
1501 };
1502
1503 static struct diff_words_style diff_words_styles[] = {
1504 { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
1505 { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
1506 { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
1507 };
1508
1509 struct diff_words_data {
1510 struct diff_words_buffer minus, plus;
1511 const char *current_plus;
1512 int last_minus;
1513 struct diff_options *opt;
1514 regex_t *word_regex;
1515 enum diff_words_type type;
1516 struct diff_words_style *style;
1517 };
1518
1519 static int fn_out_diff_words_write_helper(struct diff_options *o,
1520 struct diff_words_style_elem *st_el,
1521 const char *newline,
1522 size_t count, const char *buf)
1523 {
1524 int print = 0;
1525 struct strbuf sb = STRBUF_INIT;
1526
1527 while (count) {
1528 char *p = memchr(buf, '\n', count);
1529 if (print)
1530 strbuf_addstr(&sb, diff_line_prefix(o));
1531
1532 if (p != buf) {
1533 const char *reset = st_el->color && *st_el->color ?
1534 GIT_COLOR_RESET : NULL;
1535 if (st_el->color && *st_el->color)
1536 strbuf_addstr(&sb, st_el->color);
1537 strbuf_addstr(&sb, st_el->prefix);
1538 strbuf_add(&sb, buf, p ? p - buf : count);
1539 strbuf_addstr(&sb, st_el->suffix);
1540 if (reset)
1541 strbuf_addstr(&sb, reset);
1542 }
1543 if (!p)
1544 goto out;
1545
1546 strbuf_addstr(&sb, newline);
1547 count -= p + 1 - buf;
1548 buf = p + 1;
1549 print = 1;
1550 if (count) {
1551 emit_diff_symbol(o, DIFF_SYMBOL_WORD_DIFF,
1552 sb.buf, sb.len, 0);
1553 strbuf_reset(&sb);
1554 }
1555 }
1556
1557 out:
1558 if (sb.len)
1559 emit_diff_symbol(o, DIFF_SYMBOL_WORD_DIFF,
1560 sb.buf, sb.len, 0);
1561 strbuf_release(&sb);
1562 return 0;
1563 }
1564
1565 /*
1566 * '--color-words' algorithm can be described as:
1567 *
1568 * 1. collect the minus/plus lines of a diff hunk, divided into
1569 * minus-lines and plus-lines;
1570 *
1571 * 2. break both minus-lines and plus-lines into words and
1572 * place them into two mmfile_t with one word for each line;
1573 *
1574 * 3. use xdiff to run diff on the two mmfile_t to get the words level diff;
1575 *
1576 * And for the common parts of the both file, we output the plus side text.
1577 * diff_words->current_plus is used to trace the current position of the plus file
1578 * which printed. diff_words->last_minus is used to trace the last minus word
1579 * printed.
1580 *
1581 * For '--graph' to work with '--color-words', we need to output the graph prefix
1582 * on each line of color words output. Generally, there are two conditions on
1583 * which we should output the prefix.
1584 *
1585 * 1. diff_words->last_minus == 0 &&
1586 * diff_words->current_plus == diff_words->plus.text.ptr
1587 *
1588 * that is: the plus text must start as a new line, and if there is no minus
1589 * word printed, a graph prefix must be printed.
1590 *
1591 * 2. diff_words->current_plus > diff_words->plus.text.ptr &&
1592 * *(diff_words->current_plus - 1) == '\n'
1593 *
1594 * that is: a graph prefix must be printed following a '\n'
1595 */
1596 static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
1597 {
1598 if ((diff_words->last_minus == 0 &&
1599 diff_words->current_plus == diff_words->plus.text.ptr) ||
1600 (diff_words->current_plus > diff_words->plus.text.ptr &&
1601 *(diff_words->current_plus - 1) == '\n')) {
1602 return 1;
1603 } else {
1604 return 0;
1605 }
1606 }
1607
1608 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
1609 {
1610 struct diff_words_data *diff_words = priv;
1611 struct diff_words_style *style = diff_words->style;
1612 int minus_first, minus_len, plus_first, plus_len;
1613 const char *minus_begin, *minus_end, *plus_begin, *plus_end;
1614 struct diff_options *opt = diff_words->opt;
1615 const char *line_prefix;
1616
1617 if (line[0] != '@' || parse_hunk_header(line, len,
1618 &minus_first, &minus_len, &plus_first, &plus_len))
1619 return;
1620
1621 assert(opt);
1622 line_prefix = diff_line_prefix(opt);
1623
1624 /* POSIX requires that first be decremented by one if len == 0... */
1625 if (minus_len) {
1626 minus_begin = diff_words->minus.orig[minus_first].begin;
1627 minus_end =
1628 diff_words->minus.orig[minus_first + minus_len - 1].end;
1629 } else
1630 minus_begin = minus_end =
1631 diff_words->minus.orig[minus_first].end;
1632
1633 if (plus_len) {
1634 plus_begin = diff_words->plus.orig[plus_first].begin;
1635 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
1636 } else
1637 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
1638
1639 if (color_words_output_graph_prefix(diff_words)) {
1640 fputs(line_prefix, diff_words->opt->file);
1641 }
1642 if (diff_words->current_plus != plus_begin) {
1643 fn_out_diff_words_write_helper(diff_words->opt,
1644 &style->ctx, style->newline,
1645 plus_begin - diff_words->current_plus,
1646 diff_words->current_plus);
1647 }
1648 if (minus_begin != minus_end) {
1649 fn_out_diff_words_write_helper(diff_words->opt,
1650 &style->old_word, style->newline,
1651 minus_end - minus_begin, minus_begin);
1652 }
1653 if (plus_begin != plus_end) {
1654 fn_out_diff_words_write_helper(diff_words->opt,
1655 &style->new_word, style->newline,
1656 plus_end - plus_begin, plus_begin);
1657 }
1658
1659 diff_words->current_plus = plus_end;
1660 diff_words->last_minus = minus_first;
1661 }
1662
1663 /* This function starts looking at *begin, and returns 0 iff a word was found. */
1664 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
1665 int *begin, int *end)
1666 {
1667 if (word_regex && *begin < buffer->size) {
1668 regmatch_t match[1];
1669 if (!regexec_buf(word_regex, buffer->ptr + *begin,
1670 buffer->size - *begin, 1, match, 0)) {
1671 char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
1672 '\n', match[0].rm_eo - match[0].rm_so);
1673 *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
1674 *begin += match[0].rm_so;
1675 return *begin >= *end;
1676 }
1677 return -1;
1678 }
1679
1680 /* find the next word */
1681 while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
1682 (*begin)++;
1683 if (*begin >= buffer->size)
1684 return -1;
1685
1686 /* find the end of the word */
1687 *end = *begin + 1;
1688 while (*end < buffer->size && !isspace(buffer->ptr[*end]))
1689 (*end)++;
1690
1691 return 0;
1692 }
1693
1694 /*
1695 * This function splits the words in buffer->text, stores the list with
1696 * newline separator into out, and saves the offsets of the original words
1697 * in buffer->orig.
1698 */
1699 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
1700 regex_t *word_regex)
1701 {
1702 int i, j;
1703 long alloc = 0;
1704
1705 out->size = 0;
1706 out->ptr = NULL;
1707
1708 /* fake an empty "0th" word */
1709 ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
1710 buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
1711 buffer->orig_nr = 1;
1712
1713 for (i = 0; i < buffer->text.size; i++) {
1714 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
1715 return;
1716
1717 /* store original boundaries */
1718 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
1719 buffer->orig_alloc);
1720 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
1721 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
1722 buffer->orig_nr++;
1723
1724 /* store one word */
1725 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
1726 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
1727 out->ptr[out->size + j - i] = '\n';
1728 out->size += j - i + 1;
1729
1730 i = j - 1;
1731 }
1732 }
1733
1734 /* this executes the word diff on the accumulated buffers */
1735 static void diff_words_show(struct diff_words_data *diff_words)
1736 {
1737 xpparam_t xpp;
1738 xdemitconf_t xecfg;
1739 mmfile_t minus, plus;
1740 struct diff_words_style *style = diff_words->style;
1741
1742 struct diff_options *opt = diff_words->opt;
1743 const char *line_prefix;
1744
1745 assert(opt);
1746 line_prefix = diff_line_prefix(opt);
1747
1748 /* special case: only removal */
1749 if (!diff_words->plus.text.size) {
1750 emit_diff_symbol(diff_words->opt, DIFF_SYMBOL_WORD_DIFF,
1751 line_prefix, strlen(line_prefix), 0);
1752 fn_out_diff_words_write_helper(diff_words->opt,
1753 &style->old_word, style->newline,
1754 diff_words->minus.text.size,
1755 diff_words->minus.text.ptr);
1756 diff_words->minus.text.size = 0;
1757 return;
1758 }
1759
1760 diff_words->current_plus = diff_words->plus.text.ptr;
1761 diff_words->last_minus = 0;
1762
1763 memset(&xpp, 0, sizeof(xpp));
1764 memset(&xecfg, 0, sizeof(xecfg));
1765 diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
1766 diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
1767 xpp.flags = 0;
1768 /* as only the hunk header will be parsed, we need a 0-context */
1769 xecfg.ctxlen = 0;
1770 if (xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
1771 &xpp, &xecfg))
1772 die("unable to generate word diff");
1773 free(minus.ptr);
1774 free(plus.ptr);
1775 if (diff_words->current_plus != diff_words->plus.text.ptr +
1776 diff_words->plus.text.size) {
1777 if (color_words_output_graph_prefix(diff_words))
1778 emit_diff_symbol(diff_words->opt, DIFF_SYMBOL_WORD_DIFF,
1779 line_prefix, strlen(line_prefix), 0);
1780 fn_out_diff_words_write_helper(diff_words->opt,
1781 &style->ctx, style->newline,
1782 diff_words->plus.text.ptr + diff_words->plus.text.size
1783 - diff_words->current_plus, diff_words->current_plus);
1784 }
1785 diff_words->minus.text.size = diff_words->plus.text.size = 0;
1786 }
1787
1788 /* In "color-words" mode, show word-diff of words accumulated in the buffer */
1789 static void diff_words_flush(struct emit_callback *ecbdata)
1790 {
1791 struct diff_options *wo = ecbdata->diff_words->opt;
1792
1793 if (ecbdata->diff_words->minus.text.size ||
1794 ecbdata->diff_words->plus.text.size)
1795 diff_words_show(ecbdata->diff_words);
1796
1797 if (wo->emitted_symbols) {
1798 struct diff_options *o = ecbdata->opt;
1799 struct emitted_diff_symbols *wol = wo->emitted_symbols;
1800 int i;
1801
1802 /*
1803 * NEEDSWORK:
1804 * Instead of appending each, concat all words to a line?
1805 */
1806 for (i = 0; i < wol->nr; i++)
1807 append_emitted_diff_symbol(o, &wol->buf[i]);
1808
1809 for (i = 0; i < wol->nr; i++)
1810 free((void *)wol->buf[i].line);
1811
1812 wol->nr = 0;
1813 }
1814 }
1815
1816 static void diff_filespec_load_driver(struct diff_filespec *one)
1817 {
1818 /* Use already-loaded driver */
1819 if (one->driver)
1820 return;
1821
1822 if (S_ISREG(one->mode))
1823 one->driver = userdiff_find_by_path(one->path);
1824
1825 /* Fallback to default settings */
1826 if (!one->driver)
1827 one->driver = userdiff_find_by_name("default");
1828 }
1829
1830 static const char *userdiff_word_regex(struct diff_filespec *one)
1831 {
1832 diff_filespec_load_driver(one);
1833 return one->driver->word_regex;
1834 }
1835
1836 static void init_diff_words_data(struct emit_callback *ecbdata,
1837 struct diff_options *orig_opts,
1838 struct diff_filespec *one,
1839 struct diff_filespec *two)
1840 {
1841 int i;
1842 struct diff_options *o = xmalloc(sizeof(struct diff_options));
1843 memcpy(o, orig_opts, sizeof(struct diff_options));
1844
1845 ecbdata->diff_words =
1846 xcalloc(1, sizeof(struct diff_words_data));
1847 ecbdata->diff_words->type = o->word_diff;
1848 ecbdata->diff_words->opt = o;
1849
1850 if (orig_opts->emitted_symbols)
1851 o->emitted_symbols =
1852 xcalloc(1, sizeof(struct emitted_diff_symbols));
1853
1854 if (!o->word_regex)
1855 o->word_regex = userdiff_word_regex(one);
1856 if (!o->word_regex)
1857 o->word_regex = userdiff_word_regex(two);
1858 if (!o->word_regex)
1859 o->word_regex = diff_word_regex_cfg;
1860 if (o->word_regex) {
1861 ecbdata->diff_words->word_regex = (regex_t *)
1862 xmalloc(sizeof(regex_t));
1863 if (regcomp(ecbdata->diff_words->word_regex,
1864 o->word_regex,
1865 REG_EXTENDED | REG_NEWLINE))
1866 die("invalid regular expression: %s",
1867 o->word_regex);
1868 }
1869 for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
1870 if (o->word_diff == diff_words_styles[i].type) {
1871 ecbdata->diff_words->style =
1872 &diff_words_styles[i];
1873 break;
1874 }
1875 }
1876 if (want_color(o->use_color)) {
1877 struct diff_words_style *st = ecbdata->diff_words->style;
1878 st->old_word.color = diff_get_color_opt(o, DIFF_FILE_OLD);
1879 st->new_word.color = diff_get_color_opt(o, DIFF_FILE_NEW);
1880 st->ctx.color = diff_get_color_opt(o, DIFF_CONTEXT);
1881 }
1882 }
1883
1884 static void free_diff_words_data(struct emit_callback *ecbdata)
1885 {
1886 if (ecbdata->diff_words) {
1887 diff_words_flush(ecbdata);
1888 free (ecbdata->diff_words->opt->emitted_symbols);
1889 free (ecbdata->diff_words->opt);
1890 free (ecbdata->diff_words->minus.text.ptr);
1891 free (ecbdata->diff_words->minus.orig);
1892 free (ecbdata->diff_words->plus.text.ptr);
1893 free (ecbdata->diff_words->plus.orig);
1894 if (ecbdata->diff_words->word_regex) {
1895 regfree(ecbdata->diff_words->word_regex);
1896 free(ecbdata->diff_words->word_regex);
1897 }
1898 FREE_AND_NULL(ecbdata->diff_words);
1899 }
1900 }
1901
1902 const char *diff_get_color(int diff_use_color, enum color_diff ix)
1903 {
1904 if (want_color(diff_use_color))
1905 return diff_colors[ix];
1906 return "";
1907 }
1908
1909 const char *diff_line_prefix(struct diff_options *opt)
1910 {
1911 struct strbuf *msgbuf;
1912 if (!opt->output_prefix)
1913 return "";
1914
1915 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
1916 return msgbuf->buf;
1917 }
1918
1919 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1920 {
1921 const char *cp;
1922 unsigned long allot;
1923 size_t l = len;
1924
1925 cp = line;
1926 allot = l;
1927 while (0 < l) {
1928 (void) utf8_width(&cp, &l);
1929 if (!cp)
1930 break; /* truncated in the middle? */
1931 }
1932 return allot - l;
1933 }
1934
1935 static void find_lno(const char *line, struct emit_callback *ecbdata)
1936 {
1937 const char *p;
1938 ecbdata->lno_in_preimage = 0;
1939 ecbdata->lno_in_postimage = 0;
1940 p = strchr(line, '-');
1941 if (!p)
1942 return; /* cannot happen */
1943 ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1944 p = strchr(p, '+');
1945 if (!p)
1946 return; /* cannot happen */
1947 ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1948 }
1949
1950 static void fn_out_consume(void *priv, char *line, unsigned long len)
1951 {
1952 struct emit_callback *ecbdata = priv;
1953 const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1954 struct diff_options *o = ecbdata->opt;
1955
1956 o->found_changes = 1;
1957
1958 if (ecbdata->header) {
1959 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
1960 ecbdata->header->buf, ecbdata->header->len, 0);
1961 strbuf_reset(ecbdata->header);
1962 ecbdata->header = NULL;
1963 }
1964
1965 if (ecbdata->label_path[0]) {
1966 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_MINUS,
1967 ecbdata->label_path[0],
1968 strlen(ecbdata->label_path[0]), 0);
1969 emit_diff_symbol(o, DIFF_SYMBOL_FILEPAIR_PLUS,
1970 ecbdata->label_path[1],
1971 strlen(ecbdata->label_path[1]), 0);
1972 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1973 }
1974
1975 if (diff_suppress_blank_empty
1976 && len == 2 && line[0] == ' ' && line[1] == '\n') {
1977 line[0] = '\n';
1978 len = 1;
1979 }
1980
1981 if (line[0] == '@') {
1982 if (ecbdata->diff_words)
1983 diff_words_flush(ecbdata);
1984 len = sane_truncate_line(ecbdata, line, len);
1985 find_lno(line, ecbdata);
1986 emit_hunk_header(ecbdata, line, len);
1987 return;
1988 }
1989
1990 if (ecbdata->diff_words) {
1991 enum diff_symbol s =
1992 ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN ?
1993 DIFF_SYMBOL_WORDS_PORCELAIN : DIFF_SYMBOL_WORDS;
1994 if (line[0] == '-') {
1995 diff_words_append(line, len,
1996 &ecbdata->diff_words->minus);
1997 return;
1998 } else if (line[0] == '+') {
1999 diff_words_append(line, len,
2000 &ecbdata->diff_words->plus);
2001 return;
2002 } else if (starts_with(line, "\\ ")) {
2003 /*
2004 * Eat the "no newline at eof" marker as if we
2005 * saw a "+" or "-" line with nothing on it,
2006 * and return without diff_words_flush() to
2007 * defer processing. If this is the end of
2008 * preimage, more "+" lines may come after it.
2009 */
2010 return;
2011 }
2012 diff_words_flush(ecbdata);
2013 emit_diff_symbol(o, s, line, len, 0);
2014 return;
2015 }
2016
2017 switch (line[0]) {
2018 case '+':
2019 ecbdata->lno_in_postimage++;
2020 emit_add_line(reset, ecbdata, line + 1, len - 1);
2021 break;
2022 case '-':
2023 ecbdata->lno_in_preimage++;
2024 emit_del_line(reset, ecbdata, line + 1, len - 1);
2025 break;
2026 case ' ':
2027 ecbdata->lno_in_postimage++;
2028 ecbdata->lno_in_preimage++;
2029 emit_context_line(reset, ecbdata, line + 1, len - 1);
2030 break;
2031 default:
2032 /* incomplete line at the end */
2033 ecbdata->lno_in_preimage++;
2034 emit_diff_symbol(o, DIFF_SYMBOL_CONTEXT_INCOMPLETE,
2035 line, len, 0);
2036 break;
2037 }
2038 }
2039
2040 static void pprint_rename(struct strbuf *name, const char *a, const char *b)
2041 {
2042 const char *old_name = a;
2043 const char *new_name = b;
2044 int pfx_length, sfx_length;
2045 int pfx_adjust_for_slash;
2046 int len_a = strlen(a);
2047 int len_b = strlen(b);
2048 int a_midlen, b_midlen;
2049 int qlen_a = quote_c_style(a, NULL, NULL, 0);
2050 int qlen_b = quote_c_style(b, NULL, NULL, 0);
2051
2052 if (qlen_a || qlen_b) {
2053 quote_c_style(a, name, NULL, 0);
2054 strbuf_addstr(name, " => ");
2055 quote_c_style(b, name, NULL, 0);
2056 return;
2057 }
2058
2059 /* Find common prefix */
2060 pfx_length = 0;
2061 while (*old_name && *new_name && *old_name == *new_name) {
2062 if (*old_name == '/')
2063 pfx_length = old_name - a + 1;
2064 old_name++;
2065 new_name++;
2066 }
2067
2068 /* Find common suffix */
2069 old_name = a + len_a;
2070 new_name = b + len_b;
2071 sfx_length = 0;
2072 /*
2073 * If there is a common prefix, it must end in a slash. In
2074 * that case we let this loop run 1 into the prefix to see the
2075 * same slash.
2076 *
2077 * If there is no common prefix, we cannot do this as it would
2078 * underrun the input strings.
2079 */
2080 pfx_adjust_for_slash = (pfx_length ? 1 : 0);
2081 while (a + pfx_length - pfx_adjust_for_slash <= old_name &&
2082 b + pfx_length - pfx_adjust_for_slash <= new_name &&
2083 *old_name == *new_name) {
2084 if (*old_name == '/')
2085 sfx_length = len_a - (old_name - a);
2086 old_name--;
2087 new_name--;
2088 }
2089
2090 /*
2091 * pfx{mid-a => mid-b}sfx
2092 * {pfx-a => pfx-b}sfx
2093 * pfx{sfx-a => sfx-b}
2094 * name-a => name-b
2095 */
2096 a_midlen = len_a - pfx_length - sfx_length;
2097 b_midlen = len_b - pfx_length - sfx_length;
2098 if (a_midlen < 0)
2099 a_midlen = 0;
2100 if (b_midlen < 0)
2101 b_midlen = 0;
2102
2103 strbuf_grow(name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
2104 if (pfx_length + sfx_length) {
2105 strbuf_add(name, a, pfx_length);
2106 strbuf_addch(name, '{');
2107 }
2108 strbuf_add(name, a + pfx_length, a_midlen);
2109 strbuf_addstr(name, " => ");
2110 strbuf_add(name, b + pfx_length, b_midlen);
2111 if (pfx_length + sfx_length) {
2112 strbuf_addch(name, '}');
2113 strbuf_add(name, a + len_a - sfx_length, sfx_length);
2114 }
2115 }
2116
2117 struct diffstat_t {
2118 int nr;
2119 int alloc;
2120 struct diffstat_file {
2121 char *from_name;
2122 char *name;
2123 char *print_name;
2124 const char *comments;
2125 unsigned is_unmerged:1;
2126 unsigned is_binary:1;
2127 unsigned is_renamed:1;
2128 unsigned is_interesting:1;
2129 uintmax_t added, deleted;
2130 } **files;
2131 };
2132
2133 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
2134 const char *name_a,
2135 const char *name_b)
2136 {
2137 struct diffstat_file *x;
2138 x = xcalloc(1, sizeof(*x));
2139 ALLOC_GROW(diffstat->files, diffstat->nr + 1, diffstat->alloc);
2140 diffstat->files[diffstat->nr++] = x;
2141 if (name_b) {
2142 x->from_name = xstrdup(name_a);
2143 x->name = xstrdup(name_b);
2144 x->is_renamed = 1;
2145 }
2146 else {
2147 x->from_name = NULL;
2148 x->name = xstrdup(name_a);
2149 }
2150 return x;
2151 }
2152
2153 static void diffstat_consume(void *priv, char *line, unsigned long len)
2154 {
2155 struct diffstat_t *diffstat = priv;
2156 struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
2157
2158 if (line[0] == '+')
2159 x->added++;
2160 else if (line[0] == '-')
2161 x->deleted++;
2162 }
2163
2164 const char mime_boundary_leader[] = "------------";
2165
2166 static int scale_linear(int it, int width, int max_change)
2167 {
2168 if (!it)
2169 return 0;
2170 /*
2171 * make sure that at least one '-' or '+' is printed if
2172 * there is any change to this path. The easiest way is to
2173 * scale linearly as if the alloted width is one column shorter
2174 * than it is, and then add 1 to the result.
2175 */
2176 return 1 + (it * (width - 1) / max_change);
2177 }
2178
2179 static void show_graph(struct strbuf *out, char ch, int cnt,
2180 const char *set, const char *reset)
2181 {
2182 if (cnt <= 0)
2183 return;
2184 strbuf_addstr(out, set);
2185 strbuf_addchars(out, ch, cnt);
2186 strbuf_addstr(out, reset);
2187 }
2188
2189 static void fill_print_name(struct diffstat_file *file)
2190 {
2191 struct strbuf pname = STRBUF_INIT;
2192
2193 if (file->print_name)
2194 return;
2195
2196 if (file->is_renamed)
2197 pprint_rename(&pname, file->from_name, file->name);
2198 else
2199 quote_c_style(file->name, &pname, NULL, 0);
2200
2201 if (file->comments)
2202 strbuf_addf(&pname, " (%s)", file->comments);
2203
2204 file->print_name = strbuf_detach(&pname, NULL);
2205 }
2206
2207 static void print_stat_summary_inserts_deletes(struct diff_options *options,
2208 int files, int insertions, int deletions)
2209 {
2210 struct strbuf sb = STRBUF_INIT;
2211
2212 if (!files) {
2213 assert(insertions == 0 && deletions == 0);
2214 emit_diff_symbol(options, DIFF_SYMBOL_STATS_SUMMARY_NO_FILES,
2215 NULL, 0, 0);
2216 return;
2217 }
2218
2219 strbuf_addf(&sb,
2220 (files == 1) ? " %d file changed" : " %d files changed",
2221 files);
2222
2223 /*
2224 * For binary diff, the caller may want to print "x files
2225 * changed" with insertions == 0 && deletions == 0.
2226 *
2227 * Not omitting "0 insertions(+), 0 deletions(-)" in this case
2228 * is probably less confusing (i.e skip over "2 files changed
2229 * but nothing about added/removed lines? Is this a bug in Git?").
2230 */
2231 if (insertions || deletions == 0) {
2232 strbuf_addf(&sb,
2233 (insertions == 1) ? ", %d insertion(+)" : ", %d insertions(+)",
2234 insertions);
2235 }
2236
2237 if (deletions || insertions == 0) {
2238 strbuf_addf(&sb,
2239 (deletions == 1) ? ", %d deletion(-)" : ", %d deletions(-)",
2240 deletions);
2241 }
2242 strbuf_addch(&sb, '\n');
2243 emit_diff_symbol(options, DIFF_SYMBOL_STATS_SUMMARY_INSERTS_DELETES,
2244 sb.buf, sb.len, 0);
2245 strbuf_release(&sb);
2246 }
2247
2248 void print_stat_summary(FILE *fp, int files,
2249 int insertions, int deletions)
2250 {
2251 struct diff_options o;
2252 memset(&o, 0, sizeof(o));
2253 o.file = fp;
2254
2255 print_stat_summary_inserts_deletes(&o, files, insertions, deletions);
2256 }
2257
2258 static void show_stats(struct diffstat_t *data, struct diff_options *options)
2259 {
2260 int i, len, add, del, adds = 0, dels = 0;
2261 uintmax_t max_change = 0, max_len = 0;
2262 int total_files = data->nr, count;
2263 int width, name_width, graph_width, number_width = 0, bin_width = 0;
2264 const char *reset, *add_c, *del_c;
2265 int extra_shown = 0;
2266 const char *line_prefix = diff_line_prefix(options);
2267 struct strbuf out = STRBUF_INIT;
2268
2269 if (data->nr == 0)
2270 return;
2271
2272 count = options->stat_count ? options->stat_count : data->nr;
2273
2274 reset = diff_get_color_opt(options, DIFF_RESET);
2275 add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
2276 del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
2277
2278 /*
2279 * Find the longest filename and max number of changes
2280 */
2281 for (i = 0; (i < count) && (i < data->nr); i++) {
2282 struct diffstat_file *file = data->files[i];
2283 uintmax_t change = file->added + file->deleted;
2284
2285 if (!file->is_interesting && (change == 0)) {
2286 count++; /* not shown == room for one more */
2287 continue;
2288 }
2289 fill_print_name(file);
2290 len = strlen(file->print_name);
2291 if (max_len < len)
2292 max_len = len;
2293
2294 if (file->is_unmerged) {
2295 /* "Unmerged" is 8 characters */
2296 bin_width = bin_width < 8 ? 8 : bin_width;
2297 continue;
2298 }
2299 if (file->is_binary) {
2300 /* "Bin XXX -> YYY bytes" */
2301 int w = 14 + decimal_width(file->added)
2302 + decimal_width(file->deleted);
2303 bin_width = bin_width < w ? w : bin_width;
2304 /* Display change counts aligned with "Bin" */
2305 number_width = 3;
2306 continue;
2307 }
2308
2309 if (max_change < change)
2310 max_change = change;
2311 }
2312 count = i; /* where we can stop scanning in data->files[] */
2313
2314 /*
2315 * We have width = stat_width or term_columns() columns total.
2316 * We want a maximum of min(max_len, stat_name_width) for the name part.
2317 * We want a maximum of min(max_change, stat_graph_width) for the +- part.
2318 * We also need 1 for " " and 4 + decimal_width(max_change)
2319 * for " | NNNN " and one the empty column at the end, altogether
2320 * 6 + decimal_width(max_change).
2321 *
2322 * If there's not enough space, we will use the smaller of
2323 * stat_name_width (if set) and 5/8*width for the filename,
2324 * and the rest for constant elements + graph part, but no more
2325 * than stat_graph_width for the graph part.
2326 * (5/8 gives 50 for filename and 30 for the constant parts + graph
2327 * for the standard terminal size).
2328 *
2329 * In other words: stat_width limits the maximum width, and
2330 * stat_name_width fixes the maximum width of the filename,
2331 * and is also used to divide available columns if there
2332 * aren't enough.
2333 *
2334 * Binary files are displayed with "Bin XXX -> YYY bytes"
2335 * instead of the change count and graph. This part is treated
2336 * similarly to the graph part, except that it is not
2337 * "scaled". If total width is too small to accommodate the
2338 * guaranteed minimum width of the filename part and the
2339 * separators and this message, this message will "overflow"
2340 * making the line longer than the maximum width.
2341 */
2342
2343 if (options->stat_width == -1)
2344 width = term_columns() - strlen(line_prefix);
2345 else
2346 width = options->stat_width ? options->stat_width : 80;
2347 number_width = decimal_width(max_change) > number_width ?
2348 decimal_width(max_change) : number_width;
2349
2350 if (options->stat_graph_width == -1)
2351 options->stat_graph_width = diff_stat_graph_width;
2352
2353 /*
2354 * Guarantee 3/8*16==6 for the graph part
2355 * and 5/8*16==10 for the filename part
2356 */
2357 if (width < 16 + 6 + number_width)
2358 width = 16 + 6 + number_width;
2359
2360 /*
2361 * First assign sizes that are wanted, ignoring available width.
2362 * strlen("Bin XXX -> YYY bytes") == bin_width, and the part
2363 * starting from "XXX" should fit in graph_width.
2364 */
2365 graph_width = max_change + 4 > bin_width ? max_change : bin_width - 4;
2366 if (options->stat_graph_width &&
2367 options->stat_graph_width < graph_width)
2368 graph_width = options->stat_graph_width;
2369
2370 name_width = (options->stat_name_width > 0 &&
2371 options->stat_name_width < max_len) ?
2372 options->stat_name_width : max_len;
2373
2374 /*
2375 * Adjust adjustable widths not to exceed maximum width
2376 */
2377 if (name_width + number_width + 6 + graph_width > width) {
2378 if (graph_width > width * 3/8 - number_width - 6) {
2379 graph_width = width * 3/8 - number_width - 6;
2380 if (graph_width < 6)
2381 graph_width = 6;
2382 }
2383
2384 if (options->stat_graph_width &&
2385 graph_width > options->stat_graph_width)
2386 graph_width = options->stat_graph_width;
2387 if (name_width > width - number_width - 6 - graph_width)
2388 name_width = width - number_width - 6 - graph_width;
2389 else
2390 graph_width = width - number_width - 6 - name_width;
2391 }
2392
2393 /*
2394 * From here name_width is the width of the name area,
2395 * and graph_width is the width of the graph area.
2396 * max_change is used to scale graph properly.
2397 */
2398 for (i = 0; i < count; i++) {
2399 const char *prefix = "";
2400 struct diffstat_file *file = data->files[i];
2401 char *name = file->print_name;
2402 uintmax_t added = file->added;
2403 uintmax_t deleted = file->deleted;
2404 int name_len;
2405
2406 if (!file->is_interesting && (added + deleted == 0))
2407 continue;
2408
2409 /*
2410 * "scale" the filename
2411 */
2412 len = name_width;
2413 name_len = strlen(name);
2414 if (name_width < name_len) {
2415 char *slash;
2416 prefix = "...";
2417 len -= 3;
2418 name += name_len - len;
2419 slash = strchr(name, '/');
2420 if (slash)
2421 name = slash;
2422 }
2423
2424 if (file->is_binary) {
2425 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2426 strbuf_addf(&out, " %*s", number_width, "Bin");
2427 if (!added && !deleted) {
2428 strbuf_addch(&out, '\n');
2429 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2430 out.buf, out.len, 0);
2431 strbuf_reset(&out);
2432 continue;
2433 }
2434 strbuf_addf(&out, " %s%"PRIuMAX"%s",
2435 del_c, deleted, reset);
2436 strbuf_addstr(&out, " -> ");
2437 strbuf_addf(&out, "%s%"PRIuMAX"%s",
2438 add_c, added, reset);
2439 strbuf_addstr(&out, " bytes\n");
2440 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2441 out.buf, out.len, 0);
2442 strbuf_reset(&out);
2443 continue;
2444 }
2445 else if (file->is_unmerged) {
2446 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2447 strbuf_addstr(&out, " Unmerged\n");
2448 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2449 out.buf, out.len, 0);
2450 strbuf_reset(&out);
2451 continue;
2452 }
2453
2454 /*
2455 * scale the add/delete
2456 */
2457 add = added;
2458 del = deleted;
2459
2460 if (graph_width <= max_change) {
2461 int total = scale_linear(add + del, graph_width, max_change);
2462 if (total < 2 && add && del)
2463 /* width >= 2 due to the sanity check */
2464 total = 2;
2465 if (add < del) {
2466 add = scale_linear(add, graph_width, max_change);
2467 del = total - add;
2468 } else {
2469 del = scale_linear(del, graph_width, max_change);
2470 add = total - del;
2471 }
2472 }
2473 strbuf_addf(&out, " %s%-*s |", prefix, len, name);
2474 strbuf_addf(&out, " %*"PRIuMAX"%s",
2475 number_width, added + deleted,
2476 added + deleted ? " " : "");
2477 show_graph(&out, '+', add, add_c, reset);
2478 show_graph(&out, '-', del, del_c, reset);
2479 strbuf_addch(&out, '\n');
2480 emit_diff_symbol(options, DIFF_SYMBOL_STATS_LINE,
2481 out.buf, out.len, 0);
2482 strbuf_reset(&out);
2483 }
2484
2485 for (i = 0; i < data->nr; i++) {
2486 struct diffstat_file *file = data->files[i];
2487 uintmax_t added = file->added;
2488 uintmax_t deleted = file->deleted;
2489
2490 if (file->is_unmerged ||
2491 (!file->is_interesting && (added + deleted == 0))) {
2492 total_files--;
2493 continue;
2494 }
2495
2496 if (!file->is_binary) {
2497 adds += added;
2498 dels += deleted;
2499 }
2500 if (i < count)
2501 continue;
2502 if (!extra_shown)
2503 emit_diff_symbol(options,
2504 DIFF_SYMBOL_STATS_SUMMARY_ABBREV,
2505 NULL, 0, 0);
2506 extra_shown = 1;
2507 }
2508
2509 print_stat_summary_inserts_deletes(options, total_files, adds, dels);
2510 strbuf_release(&out);
2511 }
2512
2513 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
2514 {
2515 int i, adds = 0, dels = 0, total_files = data->nr;
2516
2517 if (data->nr == 0)
2518 return;
2519
2520 for (i = 0; i < data->nr; i++) {
2521 int added = data->files[i]->added;
2522 int deleted = data->files[i]->deleted;
2523
2524 if (data->files[i]->is_unmerged ||
2525 (!data->files[i]->is_interesting && (added + deleted == 0))) {
2526 total_files--;
2527 } else if (!data->files[i]->is_binary) { /* don't count bytes */
2528 adds += added;
2529 dels += deleted;
2530 }
2531 }
2532 print_stat_summary_inserts_deletes(options, total_files, adds, dels);
2533 }
2534
2535 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
2536 {
2537 int i;
2538
2539 if (data->nr == 0)
2540 return;
2541
2542 for (i = 0; i < data->nr; i++) {
2543 struct diffstat_file *file = data->files[i];
2544
2545 fprintf(options->file, "%s", diff_line_prefix(options));
2546
2547 if (file->is_binary)
2548 fprintf(options->file, "-\t-\t");
2549 else
2550 fprintf(options->file,
2551 "%"PRIuMAX"\t%"PRIuMAX"\t",
2552 file->added, file->deleted);
2553 if (options->line_termination) {
2554 fill_print_name(file);
2555 if (!file->is_renamed)
2556 write_name_quoted(file->name, options->file,
2557 options->line_termination);
2558 else {
2559 fputs(file->print_name, options->file);
2560 putc(options->line_termination, options->file);
2561 }
2562 } else {
2563 if (file->is_renamed) {
2564 putc('\0', options->file);
2565 write_name_quoted(file->from_name, options->file, '\0');
2566 }
2567 write_name_quoted(file->name, options->file, '\0');
2568 }
2569 }
2570 }
2571
2572 struct dirstat_file {
2573 const char *name;
2574 unsigned long changed;
2575 };
2576
2577 struct dirstat_dir {
2578 struct dirstat_file *files;
2579 int alloc, nr, permille, cumulative;
2580 };
2581
2582 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
2583 unsigned long changed, const char *base, int baselen)
2584 {
2585 unsigned long sum_changes = 0;
2586 unsigned int sources = 0;
2587 const char *line_prefix = diff_line_prefix(opt);
2588
2589 while (dir->nr) {
2590 struct dirstat_file *f = dir->files;
2591 int namelen = strlen(f->name);
2592 unsigned long changes;
2593 char *slash;
2594
2595 if (namelen < baselen)
2596 break;
2597 if (memcmp(f->name, base, baselen))
2598 break;
2599 slash = strchr(f->name + baselen, '/');
2600 if (slash) {
2601 int newbaselen = slash + 1 - f->name;
2602 changes = gather_dirstat(opt, dir, changed, f->name, newbaselen);
2603 sources++;
2604 } else {
2605 changes = f->changed;
2606 dir->files++;
2607 dir->nr--;
2608 sources += 2;
2609 }
2610 sum_changes += changes;
2611 }
2612
2613 /*
2614 * We don't report dirstat's for
2615 * - the top level
2616 * - or cases where everything came from a single directory
2617 * under this directory (sources == 1).
2618 */
2619 if (baselen && sources != 1) {
2620 if (sum_changes) {
2621 int permille = sum_changes * 1000 / changed;
2622 if (permille >= dir->permille) {
2623 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
2624 permille / 10, permille % 10, baselen, base);
2625 if (!dir->cumulative)
2626 return 0;
2627 }
2628 }
2629 }
2630 return sum_changes;
2631 }
2632
2633 static int dirstat_compare(const void *_a, const void *_b)
2634 {
2635 const struct dirstat_file *a = _a;
2636 const struct dirstat_file *b = _b;
2637 return strcmp(a->name, b->name);
2638 }
2639
2640 static void show_dirstat(struct diff_options *options)
2641 {
2642 int i;
2643 unsigned long changed;
2644 struct dirstat_dir dir;
2645 struct diff_queue_struct *q = &diff_queued_diff;
2646
2647 dir.files = NULL;
2648 dir.alloc = 0;
2649 dir.nr = 0;
2650 dir.permille = options->dirstat_permille;
2651 dir.cumulative = options->flags.dirstat_cumulative;
2652
2653 changed = 0;
2654 for (i = 0; i < q->nr; i++) {
2655 struct diff_filepair *p = q->queue[i];
2656 const char *name;
2657 unsigned long copied, added, damage;
2658 int content_changed;
2659
2660 name = p->two->path ? p->two->path : p->one->path;
2661
2662 if (p->one->oid_valid && p->two->oid_valid)
2663 content_changed = oidcmp(&p->one->oid, &p->two->oid);
2664 else
2665 content_changed = 1;
2666
2667 if (!content_changed) {
2668 /*
2669 * The SHA1 has not changed, so pre-/post-content is
2670 * identical. We can therefore skip looking at the
2671 * file contents altogether.
2672 */
2673 damage = 0;
2674 goto found_damage;
2675 }
2676
2677 if (options->flags.dirstat_by_file) {
2678 /*
2679 * In --dirstat-by-file mode, we don't really need to
2680 * look at the actual file contents at all.
2681 * The fact that the SHA1 changed is enough for us to
2682 * add this file to the list of results
2683 * (with each file contributing equal damage).
2684 */
2685 damage = 1;
2686 goto found_damage;
2687 }
2688
2689 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
2690 diff_populate_filespec(p->one, 0);
2691 diff_populate_filespec(p->two, 0);
2692 diffcore_count_changes(p->one, p->two, NULL, NULL,
2693 &copied, &added);
2694 diff_free_filespec_data(p->one);
2695 diff_free_filespec_data(p->two);
2696 } else if (DIFF_FILE_VALID(p->one)) {
2697 diff_populate_filespec(p->one, CHECK_SIZE_ONLY);
2698 copied = added = 0;
2699 diff_free_filespec_data(p->one);
2700 } else if (DIFF_FILE_VALID(p->two)) {
2701 diff_populate_filespec(p->two, CHECK_SIZE_ONLY);
2702 copied = 0;
2703 added = p->two->size;
2704 diff_free_filespec_data(p->two);
2705 } else
2706 continue;
2707
2708 /*
2709 * Original minus copied is the removed material,
2710 * added is the new material. They are both damages
2711 * made to the preimage.
2712 * If the resulting damage is zero, we know that
2713 * diffcore_count_changes() considers the two entries to
2714 * be identical, but since content_changed is true, we
2715 * know that there must have been _some_ kind of change,
2716 * so we force all entries to have damage > 0.
2717 */
2718 damage = (p->one->size - copied) + added;
2719 if (!damage)
2720 damage = 1;
2721
2722 found_damage:
2723 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2724 dir.files[dir.nr].name = name;
2725 dir.files[dir.nr].changed = damage;
2726 changed += damage;
2727 dir.nr++;
2728 }
2729
2730 /* This can happen even with many files, if everything was renames */
2731 if (!changed)
2732 return;
2733
2734 /* Show all directories with more than x% of the changes */
2735 QSORT(dir.files, dir.nr, dirstat_compare);
2736 gather_dirstat(options, &dir, changed, "", 0);
2737 }
2738
2739 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
2740 {
2741 int i;
2742 unsigned long changed;
2743 struct dirstat_dir dir;
2744
2745 if (data->nr == 0)
2746 return;
2747
2748 dir.files = NULL;
2749 dir.alloc = 0;
2750 dir.nr = 0;
2751 dir.permille = options->dirstat_permille;
2752 dir.cumulative = options->flags.dirstat_cumulative;
2753
2754 changed = 0;
2755 for (i = 0; i < data->nr; i++) {
2756 struct diffstat_file *file = data->files[i];
2757 unsigned long damage = file->added + file->deleted;
2758 if (file->is_binary)
2759 /*
2760 * binary files counts bytes, not lines. Must find some
2761 * way to normalize binary bytes vs. textual lines.
2762 * The following heuristic assumes that there are 64
2763 * bytes per "line".
2764 * This is stupid and ugly, but very cheap...
2765 */
2766 damage = DIV_ROUND_UP(damage, 64);
2767 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
2768 dir.files[dir.nr].name = file->name;
2769 dir.files[dir.nr].changed = damage;
2770 changed += damage;
2771 dir.nr++;
2772 }
2773
2774 /* This can happen even with many files, if everything was renames */
2775 if (!changed)
2776 return;
2777
2778 /* Show all directories with more than x% of the changes */
2779 QSORT(dir.files, dir.nr, dirstat_compare);
2780 gather_dirstat(options, &dir, changed, "", 0);
2781 }
2782
2783 static void free_diffstat_info(struct diffstat_t *diffstat)
2784 {
2785 int i;
2786 for (i = 0; i < diffstat->nr; i++) {
2787 struct diffstat_file *f = diffstat->files[i];
2788 free(f->print_name);
2789 free(f->name);
2790 free(f->from_name);
2791 free(f);
2792 }
2793 free(diffstat->files);
2794 }
2795
2796 struct checkdiff_t {
2797 const char *filename;
2798 int lineno;
2799 int conflict_marker_size;
2800 struct diff_options *o;
2801 unsigned ws_rule;
2802 unsigned status;
2803 };
2804
2805 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
2806 {
2807 char firstchar;
2808 int cnt;
2809
2810 if (len < marker_size + 1)
2811 return 0;
2812 firstchar = line[0];
2813 switch (firstchar) {
2814 case '=': case '>': case '<': case '|':
2815 break;
2816 default:
2817 return 0;
2818 }
2819 for (cnt = 1; cnt < marker_size; cnt++)
2820 if (line[cnt] != firstchar)
2821 return 0;
2822 /* line[1] thru line[marker_size-1] are same as firstchar */
2823 if (len < marker_size + 1 || !isspace(line[marker_size]))
2824 return 0;
2825 return 1;
2826 }
2827
2828 static void checkdiff_consume(void *priv, char *line, unsigned long len)
2829 {
2830 struct checkdiff_t *data = priv;
2831 int marker_size = data->conflict_marker_size;
2832 const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
2833 const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
2834 const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
2835 char *err;
2836 const char *line_prefix;
2837
2838 assert(data->o);
2839 line_prefix = diff_line_prefix(data->o);
2840
2841 if (line[0] == '+') {
2842 unsigned bad;
2843 data->lineno++;
2844 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
2845 data->status |= 1;
2846 fprintf(data->o->file,
2847 "%s%s:%d: leftover conflict marker\n",
2848 line_prefix, data->filename, data->lineno);
2849 }
2850 bad = ws_check(line + 1, len - 1, data->ws_rule);
2851 if (!bad)
2852 return;
2853 data->status |= bad;
2854 err = whitespace_error_string(bad);
2855 fprintf(data->o->file, "%s%s:%d: %s.\n",
2856 line_prefix, data->filename, data->lineno, err);
2857 free(err);
2858 emit_line(data->o, set, reset, line, 1);
2859 ws_check_emit(line + 1, len - 1, data->ws_rule,
2860 data->o->file, set, reset, ws);
2861 } else if (line[0] == ' ') {
2862 data->lineno++;
2863 } else if (line[0] == '@') {
2864 char *plus = strchr(line, '+');
2865 if (plus)
2866 data->lineno = strtol(plus, NULL, 10) - 1;
2867 else
2868 die("invalid diff");
2869 }
2870 }
2871
2872 static unsigned char *deflate_it(char *data,
2873 unsigned long size,
2874 unsigned long *result_size)
2875 {
2876 int bound;
2877 unsigned char *deflated;
2878 git_zstream stream;
2879
2880 git_deflate_init(&stream, zlib_compression_level);
2881 bound = git_deflate_bound(&stream, size);
2882 deflated = xmalloc(bound);
2883 stream.next_out = deflated;
2884 stream.avail_out = bound;
2885
2886 stream.next_in = (unsigned char *)data;
2887 stream.avail_in = size;
2888 while (git_deflate(&stream, Z_FINISH) == Z_OK)
2889 ; /* nothing */
2890 git_deflate_end(&stream);
2891 *result_size = stream.total_out;
2892 return deflated;
2893 }
2894
2895 static void emit_binary_diff_body(struct diff_options *o,
2896 mmfile_t *one, mmfile_t *two)
2897 {
2898 void *cp;
2899 void *delta;
2900 void *deflated;
2901 void *data;
2902 unsigned long orig_size;
2903 unsigned long delta_size;
2904 unsigned long deflate_size;
2905 unsigned long data_size;
2906
2907 /* We could do deflated delta, or we could do just deflated two,
2908 * whichever is smaller.
2909 */
2910 delta = NULL;
2911 deflated = deflate_it(two->ptr, two->size, &deflate_size);
2912 if (one->size && two->size) {
2913 delta = diff_delta(one->ptr, one->size,
2914 two->ptr, two->size,
2915 &delta_size, deflate_size);
2916 if (delta) {
2917 void *to_free = delta;
2918 orig_size = delta_size;
2919 delta = deflate_it(delta, delta_size, &delta_size);
2920 free(to_free);
2921 }
2922 }
2923
2924 if (delta && delta_size < deflate_size) {
2925 char *s = xstrfmt("%lu", orig_size);
2926 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER_DELTA,
2927 s, strlen(s), 0);
2928 free(s);
2929 free(deflated);
2930 data = delta;
2931 data_size = delta_size;
2932 } else {
2933 char *s = xstrfmt("%lu", two->size);
2934 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER_LITERAL,
2935 s, strlen(s), 0);
2936 free(s);
2937 free(delta);
2938 data = deflated;
2939 data_size = deflate_size;
2940 }
2941
2942 /* emit data encoded in base85 */
2943 cp = data;
2944 while (data_size) {
2945 int len;
2946 int bytes = (52 < data_size) ? 52 : data_size;
2947 char line[71];
2948 data_size -= bytes;
2949 if (bytes <= 26)
2950 line[0] = bytes + 'A' - 1;
2951 else
2952 line[0] = bytes - 26 + 'a' - 1;
2953 encode_85(line + 1, cp, bytes);
2954 cp = (char *) cp + bytes;
2955
2956 len = strlen(line);
2957 line[len++] = '\n';
2958 line[len] = '\0';
2959
2960 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_BODY,
2961 line, len, 0);
2962 }
2963 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_FOOTER, NULL, 0, 0);
2964 free(data);
2965 }
2966
2967 static void emit_binary_diff(struct diff_options *o,
2968 mmfile_t *one, mmfile_t *two)
2969 {
2970 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_DIFF_HEADER, NULL, 0, 0);
2971 emit_binary_diff_body(o, one, two);
2972 emit_binary_diff_body(o, two, one);
2973 }
2974
2975 int diff_filespec_is_binary(struct diff_filespec *one)
2976 {
2977 if (one->is_binary == -1) {
2978 diff_filespec_load_driver(one);
2979 if (one->driver->binary != -1)
2980 one->is_binary = one->driver->binary;
2981 else {
2982 if (!one->data && DIFF_FILE_VALID(one))
2983 diff_populate_filespec(one, CHECK_BINARY);
2984 if (one->is_binary == -1 && one->data)
2985 one->is_binary = buffer_is_binary(one->data,
2986 one->size);
2987 if (one->is_binary == -1)
2988 one->is_binary = 0;
2989 }
2990 }
2991 return one->is_binary;
2992 }
2993
2994 static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
2995 {
2996 diff_filespec_load_driver(one);
2997 return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
2998 }
2999
3000 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
3001 {
3002 if (!options->a_prefix)
3003 options->a_prefix = a;
3004 if (!options->b_prefix)
3005 options->b_prefix = b;
3006 }
3007
3008 struct userdiff_driver *get_textconv(struct diff_filespec *one)
3009 {
3010 if (!DIFF_FILE_VALID(one))
3011 return NULL;
3012
3013 diff_filespec_load_driver(one);
3014 return userdiff_get_textconv(one->driver);
3015 }
3016
3017 static void builtin_diff(const char *name_a,
3018 const char *name_b,
3019 struct diff_filespec *one,
3020 struct diff_filespec *two,
3021 const char *xfrm_msg,
3022 int must_show_header,
3023 struct diff_options *o,
3024 int complete_rewrite)
3025 {
3026 mmfile_t mf1, mf2;
3027 const char *lbl[2];
3028 char *a_one, *b_two;
3029 const char *meta = diff_get_color_opt(o, DIFF_METAINFO);
3030 const char *reset = diff_get_color_opt(o, DIFF_RESET);
3031 const char *a_prefix, *b_prefix;
3032 struct userdiff_driver *textconv_one = NULL;
3033 struct userdiff_driver *textconv_two = NULL;
3034 struct strbuf header = STRBUF_INIT;
3035 const char *line_prefix = diff_line_prefix(o);
3036
3037 diff_set_mnemonic_prefix(o, "a/", "b/");
3038 if (o->flags.reverse_diff) {
3039 a_prefix = o->b_prefix;
3040 b_prefix = o->a_prefix;
3041 } else {
3042 a_prefix = o->a_prefix;
3043 b_prefix = o->b_prefix;
3044 }
3045
3046 if (o->submodule_format == DIFF_SUBMODULE_LOG &&
3047 (!one->mode || S_ISGITLINK(one->mode)) &&
3048 (!two->mode || S_ISGITLINK(two->mode))) {
3049 show_submodule_summary(o, one->path ? one->path : two->path,
3050 &one->oid, &two->oid,
3051 two->dirty_submodule);
3052 return;
3053 } else if (o->submodule_format == DIFF_SUBMODULE_INLINE_DIFF &&
3054 (!one->mode || S_ISGITLINK(one->mode)) &&
3055 (!two->mode || S_ISGITLINK(two->mode))) {
3056 show_submodule_inline_diff(o, one->path ? one->path : two->path,
3057 &one->oid, &two->oid,
3058 two->dirty_submodule);
3059 return;
3060 }
3061
3062 if (o->flags.allow_textconv) {
3063 textconv_one = get_textconv(one);
3064 textconv_two = get_textconv(two);
3065 }
3066
3067 /* Never use a non-valid filename anywhere if at all possible */
3068 name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
3069 name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
3070
3071 a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
3072 b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
3073 lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
3074 lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
3075 strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, meta, a_one, b_two, reset);
3076 if (lbl[0][0] == '/') {
3077 /* /dev/null */
3078 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, meta, two->mode, reset);
3079 if (xfrm_msg)
3080 strbuf_addstr(&header, xfrm_msg);
3081 must_show_header = 1;
3082 }
3083 else if (lbl[1][0] == '/') {
3084 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, meta, one->mode, reset);
3085 if (xfrm_msg)
3086 strbuf_addstr(&header, xfrm_msg);
3087 must_show_header = 1;
3088 }
3089 else {
3090 if (one->mode != two->mode) {
3091 strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, meta, one->mode, reset);
3092 strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, meta, two->mode, reset);
3093 must_show_header = 1;
3094 }
3095 if (xfrm_msg)
3096 strbuf_addstr(&header, xfrm_msg);
3097
3098 /*
3099 * we do not run diff between different kind
3100 * of objects.
3101 */
3102 if ((one->mode ^ two->mode) & S_IFMT)
3103 goto free_ab_and_return;
3104 if (complete_rewrite &&
3105 (textconv_one || !diff_filespec_is_binary(one)) &&
3106 (textconv_two || !diff_filespec_is_binary(two))) {
3107 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3108 header.buf, header.len, 0);
3109 strbuf_reset(&header);
3110 emit_rewrite_diff(name_a, name_b, one, two,
3111 textconv_one, textconv_two, o);
3112 o->found_changes = 1;
3113 goto free_ab_and_return;
3114 }
3115 }
3116
3117 if (o->irreversible_delete && lbl[1][0] == '/') {
3118 emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf,
3119 header.len, 0);
3120 strbuf_reset(&header);
3121 goto free_ab_and_return;
3122 } else if (!o->flags.text &&
3123 ( (!textconv_one && diff_filespec_is_binary(one)) ||
3124 (!textconv_two && diff_filespec_is_binary(two)) )) {
3125 struct strbuf sb = STRBUF_INIT;
3126 if (!one->data && !two->data &&
3127 S_ISREG(one->mode) && S_ISREG(two->mode) &&
3128 !o->flags.binary) {
3129 if (!oidcmp(&one->oid, &two->oid)) {
3130 if (must_show_header)
3131 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3132 header.buf, header.len,
3133 0);
3134 goto free_ab_and_return;
3135 }
3136 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3137 header.buf, header.len, 0);
3138 strbuf_addf(&sb, "%sBinary files %s and %s differ\n",
3139 diff_line_prefix(o), lbl[0], lbl[1]);
3140 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_FILES,
3141 sb.buf, sb.len, 0);
3142 strbuf_release(&sb);
3143 goto free_ab_and_return;
3144 }
3145 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
3146 die("unable to read files to diff");
3147 /* Quite common confusing case */
3148 if (mf1.size == mf2.size &&
3149 !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
3150 if (must_show_header)
3151 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3152 header.buf, header.len, 0);
3153 goto free_ab_and_return;
3154 }
3155 emit_diff_symbol(o, DIFF_SYMBOL_HEADER, header.buf, header.len, 0);
3156 strbuf_reset(&header);
3157 if (o->flags.binary)
3158 emit_binary_diff(o, &mf1, &mf2);
3159 else {
3160 strbuf_addf(&sb, "%sBinary files %s and %s differ\n",
3161 diff_line_prefix(o), lbl[0], lbl[1]);
3162 emit_diff_symbol(o, DIFF_SYMBOL_BINARY_FILES,
3163 sb.buf, sb.len, 0);
3164 strbuf_release(&sb);
3165 }
3166 o->found_changes = 1;
3167 } else {
3168 /* Crazy xdl interfaces.. */
3169 const char *diffopts = getenv("GIT_DIFF_OPTS");
3170 const char *v;
3171 xpparam_t xpp;
3172 xdemitconf_t xecfg;
3173 struct emit_callback ecbdata;
3174 const struct userdiff_funcname *pe;
3175
3176 if (must_show_header) {
3177 emit_diff_symbol(o, DIFF_SYMBOL_HEADER,
3178 header.buf, header.len, 0);
3179 strbuf_reset(&header);
3180 }
3181
3182 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
3183 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
3184
3185 pe = diff_funcname_pattern(one);
3186 if (!pe)
3187 pe = diff_funcname_pattern(two);
3188
3189 memset(&xpp, 0, sizeof(xpp));
3190 memset(&xecfg, 0, sizeof(xecfg));
3191 memset(&ecbdata, 0, sizeof(ecbdata));
3192 ecbdata.label_path = lbl;
3193 ecbdata.color_diff = want_color(o->use_color);
3194 ecbdata.ws_rule = whitespace_rule(name_b);
3195 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
3196 check_blank_at_eof(&mf1, &mf2, &ecbdata);
3197 ecbdata.opt = o;
3198 ecbdata.header = header.len ? &header : NULL;
3199 xpp.flags = o->xdl_opts;
3200 xpp.anchors = o->anchors;
3201 xpp.anchors_nr = o->anchors_nr;
3202 xecfg.ctxlen = o->context;
3203 xecfg.interhunkctxlen = o->interhunkcontext;
3204 xecfg.flags = XDL_EMIT_FUNCNAMES;
3205 if (o->flags.funccontext)
3206 xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
3207 if (pe)
3208 xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
3209 if (!diffopts)
3210 ;
3211 else if (skip_prefix(diffopts, "--unified=", &v))
3212 xecfg.ctxlen = strtoul(v, NULL, 10);
3213 else if (skip_prefix(diffopts, "-u", &v))
3214 xecfg.ctxlen = strtoul(v, NULL, 10);
3215 if (o->word_diff)
3216 init_diff_words_data(&ecbdata, o, one, two);
3217 if (xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
3218 &xpp, &xecfg))
3219 die("unable to generate diff for %s", one->path);
3220 if (o->word_diff)
3221 free_diff_words_data(&ecbdata);
3222 if (textconv_one)
3223 free(mf1.ptr);
3224 if (textconv_two)
3225 free(mf2.ptr);
3226 xdiff_clear_find_func(&xecfg);
3227 }
3228
3229 free_ab_and_return:
3230 strbuf_release(&header);
3231 diff_free_filespec_data(one);
3232 diff_free_filespec_data(two);
3233 free(a_one);
3234 free(b_two);
3235 return;
3236 }
3237
3238 static char *get_compact_summary(const struct diff_filepair *p, int is_renamed)
3239 {
3240 if (!is_renamed) {
3241 if (p->status == DIFF_STATUS_ADDED) {
3242 if (S_ISLNK(p->two->mode))
3243 return "new +l";
3244 else if ((p->two->mode & 0777) == 0755)
3245 return "new +x";
3246 else
3247 return "new";
3248 } else if (p->status == DIFF_STATUS_DELETED)
3249 return "gone";
3250 }
3251 if (S_ISLNK(p->one->mode) && !S_ISLNK(p->two->mode))
3252 return "mode -l";
3253 else if (!S_ISLNK(p->one->mode) && S_ISLNK(p->two->mode))
3254 return "mode +l";
3255 else if ((p->one->mode & 0777) == 0644 &&
3256 (p->two->mode & 0777) == 0755)
3257 return "mode +x";
3258 else if ((p->one->mode & 0777) == 0755 &&
3259 (p->two->mode & 0777) == 0644)
3260 return "mode -x";
3261 return NULL;
3262 }
3263
3264 static void builtin_diffstat(const char *name_a, const char *name_b,
3265 struct diff_filespec *one,
3266 struct diff_filespec *two,
3267 struct diffstat_t *diffstat,
3268 struct diff_options *o,
3269 struct diff_filepair *p)
3270 {
3271 mmfile_t mf1, mf2;
3272 struct diffstat_file *data;
3273 int same_contents;
3274 int complete_rewrite = 0;
3275
3276 if (!DIFF_PAIR_UNMERGED(p)) {
3277 if (p->status == DIFF_STATUS_MODIFIED && p->score)
3278 complete_rewrite = 1;
3279 }
3280
3281 data = diffstat_add(diffstat, name_a, name_b);
3282 data->is_interesting = p->status != DIFF_STATUS_UNKNOWN;
3283 if (o->flags.stat_with_summary)
3284 data->comments = get_compact_summary(p, data->is_renamed);
3285
3286 if (!one || !two) {
3287 data->is_unmerged = 1;
3288 return;
3289 }
3290
3291 same_contents = !oidcmp(&one->oid, &two->oid);
3292
3293 if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
3294 data->is_binary = 1;
3295 if (same_contents) {
3296 data->added = 0;
3297 data->deleted = 0;
3298 } else {
3299 data->added = diff_filespec_size(two);
3300 data->deleted = diff_filespec_size(one);
3301 }
3302 }
3303
3304 else if (complete_rewrite) {
3305 diff_populate_filespec(one, 0);
3306 diff_populate_filespec(two, 0);
3307 data->deleted = count_lines(one->data, one->size);
3308 data->added = count_lines(two->data, two->size);
3309 }
3310
3311 else if (!same_contents) {
3312 /* Crazy xdl interfaces.. */
3313 xpparam_t xpp;
3314 xdemitconf_t xecfg;
3315
3316 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
3317 die("unable to read files to diff");
3318
3319 memset(&xpp, 0, sizeof(xpp));
3320 memset(&xecfg, 0, sizeof(xecfg));
3321 xpp.flags = o->xdl_opts;
3322 xpp.anchors = o->anchors;
3323 xpp.anchors_nr = o->anchors_nr;
3324 xecfg.ctxlen = o->context;
3325 xecfg.interhunkctxlen = o->interhunkcontext;
3326 if (xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
3327 &xpp, &xecfg))
3328 die("unable to generate diffstat for %s", one->path);
3329 }
3330
3331 diff_free_filespec_data(one);
3332 diff_free_filespec_data(two);
3333 }
3334
3335 static void builtin_checkdiff(const char *name_a, const char *name_b,
3336 const char *attr_path,
3337 struct diff_filespec *one,
3338 struct diff_filespec *two,
3339 struct diff_options *o)
3340 {
3341 mmfile_t mf1, mf2;
3342 struct checkdiff_t data;
3343
3344 if (!two)
3345 return;
3346
3347 memset(&data, 0, sizeof(data));
3348 data.filename = name_b ? name_b : name_a;
3349 data.lineno = 0;
3350 data.o = o;
3351 data.ws_rule = whitespace_rule(attr_path);
3352 data.conflict_marker_size = ll_merge_marker_size(attr_path);
3353
3354 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
3355 die("unable to read files to diff");
3356
3357 /*
3358 * All the other codepaths check both sides, but not checking
3359 * the "old" side here is deliberate. We are checking the newly
3360 * introduced changes, and as long as the "new" side is text, we
3361 * can and should check what it introduces.
3362 */
3363 if (diff_filespec_is_binary(two))
3364 goto free_and_return;
3365 else {
3366 /* Crazy xdl interfaces.. */
3367 xpparam_t xpp;
3368 xdemitconf_t xecfg;
3369
3370 memset(&xpp, 0, sizeof(xpp));
3371 memset(&xecfg, 0, sizeof(xecfg));
3372 xecfg.ctxlen = 1; /* at least one context line */
3373 xpp.flags = 0;
3374 if (xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
3375 &xpp, &xecfg))
3376 die("unable to generate checkdiff for %s", one->path);
3377
3378 if (data.ws_rule & WS_BLANK_AT_EOF) {
3379 struct emit_callback ecbdata;
3380 int blank_at_eof;
3381
3382 ecbdata.ws_rule = data.ws_rule;
3383 check_blank_at_eof(&mf1, &mf2, &ecbdata);
3384 blank_at_eof = ecbdata.blank_at_eof_in_postimage;
3385
3386 if (blank_at_eof) {
3387 static char *err;
3388 if (!err)
3389 err = whitespace_error_string(WS_BLANK_AT_EOF);
3390 fprintf(o->file, "%s:%d: %s.\n",
3391 data.filename, blank_at_eof, err);
3392 data.status = 1; /* report errors */
3393 }
3394 }
3395 }
3396 free_and_return:
3397 diff_free_filespec_data(one);
3398 diff_free_filespec_data(two);
3399 if (data.status)
3400 o->flags.check_failed = 1;
3401 }
3402
3403 struct diff_filespec *alloc_filespec(const char *path)
3404 {
3405 struct diff_filespec *spec;
3406
3407 FLEXPTR_ALLOC_STR(spec, path, path);
3408 spec->count = 1;
3409 spec->is_binary = -1;
3410 return spec;
3411 }
3412
3413 void free_filespec(struct diff_filespec *spec)
3414 {
3415 if (!--spec->count) {
3416 diff_free_filespec_data(spec);
3417 free(spec);
3418 }
3419 }
3420
3421 void fill_filespec(struct diff_filespec *spec, const struct object_id *oid,
3422 int oid_valid, unsigned short mode)
3423 {
3424 if (mode) {
3425 spec->mode = canon_mode(mode);
3426 oidcpy(&spec->oid, oid);
3427 spec->oid_valid = oid_valid;
3428 }
3429 }
3430
3431 /*
3432 * Given a name and sha1 pair, if the index tells us the file in
3433 * the work tree has that object contents, return true, so that
3434 * prepare_temp_file() does not have to inflate and extract.
3435 */
3436 static int reuse_worktree_file(const char *name, const struct object_id *oid, int want_file)
3437 {
3438 const struct cache_entry *ce;
3439 struct stat st;
3440 int pos, len;
3441
3442 /*
3443 * We do not read the cache ourselves here, because the
3444 * benchmark with my previous version that always reads cache
3445 * shows that it makes things worse for diff-tree comparing
3446 * two linux-2.6 kernel trees in an already checked out work
3447 * tree. This is because most diff-tree comparisons deal with
3448 * only a small number of files, while reading the cache is
3449 * expensive for a large project, and its cost outweighs the
3450 * savings we get by not inflating the object to a temporary
3451 * file. Practically, this code only helps when we are used
3452 * by diff-cache --cached, which does read the cache before
3453 * calling us.
3454 */
3455 if (!active_cache)
3456 return 0;
3457
3458 /* We want to avoid the working directory if our caller
3459 * doesn't need the data in a normal file, this system
3460 * is rather slow with its stat/open/mmap/close syscalls,
3461 * and the object is contained in a pack file. The pack
3462 * is probably already open and will be faster to obtain
3463 * the data through than the working directory. Loose
3464 * objects however would tend to be slower as they need
3465 * to be individually opened and inflated.
3466 */
3467 if (!FAST_WORKING_DIRECTORY && !want_file && has_object_pack(oid))
3468 return 0;
3469
3470 /*
3471 * Similarly, if we'd have to convert the file contents anyway, that
3472 * makes the optimization not worthwhile.
3473 */
3474 if (!want_file && would_convert_to_git(&the_index, name))
3475 return 0;
3476
3477 len = strlen(name);
3478 pos = cache_name_pos(name, len);
3479 if (pos < 0)
3480 return 0;
3481 ce = active_cache[pos];
3482
3483 /*
3484 * This is not the sha1 we are looking for, or
3485 * unreusable because it is not a regular file.
3486 */
3487 if (oidcmp(oid, &ce->oid) || !S_ISREG(ce->ce_mode))
3488 return 0;
3489
3490 /*
3491 * If ce is marked as "assume unchanged", there is no
3492 * guarantee that work tree matches what we are looking for.
3493 */
3494 if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
3495 return 0;
3496
3497 /*
3498 * If ce matches the file in the work tree, we can reuse it.
3499 */
3500 if (ce_uptodate(ce) ||
3501 (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
3502 return 1;
3503
3504 return 0;
3505 }
3506
3507 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
3508 {
3509 struct strbuf buf = STRBUF_INIT;
3510 char *dirty = "";
3511
3512 /* Are we looking at the work tree? */
3513 if (s->dirty_submodule)
3514 dirty = "-dirty";
3515
3516 strbuf_addf(&buf, "Subproject commit %s%s\n",
3517 oid_to_hex(&s->oid), dirty);
3518 s->size = buf.len;
3519 if (size_only) {
3520 s->data = NULL;
3521 strbuf_release(&buf);
3522 } else {
3523 s->data = strbuf_detach(&buf, NULL);
3524 s->should_free = 1;
3525 }
3526 return 0;
3527 }
3528
3529 /*
3530 * While doing rename detection and pickaxe operation, we may need to
3531 * grab the data for the blob (or file) for our own in-core comparison.
3532 * diff_filespec has data and size fields for this purpose.
3533 */
3534 int diff_populate_filespec(struct diff_filespec *s, unsigned int flags)
3535 {
3536 int size_only = flags & CHECK_SIZE_ONLY;
3537 int err = 0;
3538 int conv_flags = global_conv_flags_eol;
3539 /*
3540 * demote FAIL to WARN to allow inspecting the situation
3541 * instead of refusing.
3542 */
3543 if (conv_flags & CONV_EOL_RNDTRP_DIE)
3544 conv_flags = CONV_EOL_RNDTRP_WARN;
3545
3546 if (!DIFF_FILE_VALID(s))
3547 die("internal error: asking to populate invalid file.");
3548 if (S_ISDIR(s->mode))
3549 return -1;
3550
3551 if (s->data)
3552 return 0;
3553
3554 if (size_only && 0 < s->size)
3555 return 0;
3556
3557 if (S_ISGITLINK(s->mode))
3558 return diff_populate_gitlink(s, size_only);
3559
3560 if (!s->oid_valid ||
3561 reuse_worktree_file(s->path, &s->oid, 0)) {
3562 struct strbuf buf = STRBUF_INIT;
3563 struct stat st;
3564 int fd;
3565
3566 if (lstat(s->path, &st) < 0) {
3567 err_empty:
3568 err = -1;
3569 empty:
3570 s->data = (char *)"";
3571 s->size = 0;
3572 return err;
3573 }
3574 s->size = xsize_t(st.st_size);
3575 if (!s->size)
3576 goto empty;
3577 if (S_ISLNK(st.st_mode)) {
3578 struct strbuf sb = STRBUF_INIT;
3579
3580 if (strbuf_readlink(&sb, s->path, s->size))
3581 goto err_empty;
3582 s->size = sb.len;
3583 s->data = strbuf_detach(&sb, NULL);
3584 s->should_free = 1;
3585 return 0;
3586 }
3587
3588 /*
3589 * Even if the caller would be happy with getting
3590 * only the size, we cannot return early at this
3591 * point if the path requires us to run the content
3592 * conversion.
3593 */
3594 if (size_only && !would_convert_to_git(&the_index, s->path))
3595 return 0;
3596
3597 /*
3598 * Note: this check uses xsize_t(st.st_size) that may
3599 * not be the true size of the blob after it goes
3600 * through convert_to_git(). This may not strictly be
3601 * correct, but the whole point of big_file_threshold
3602 * and is_binary check being that we want to avoid
3603 * opening the file and inspecting the contents, this
3604 * is probably fine.
3605 */
3606 if ((flags & CHECK_BINARY) &&
3607 s->size > big_file_threshold && s->is_binary == -1) {
3608 s->is_binary = 1;
3609 return 0;
3610 }
3611 fd = open(s->path, O_RDONLY);
3612 if (fd < 0)
3613 goto err_empty;
3614 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
3615 close(fd);
3616 s->should_munmap = 1;
3617
3618 /*
3619 * Convert from working tree format to canonical git format
3620 */
3621 if (convert_to_git(&the_index, s->path, s->data, s->size, &buf, conv_flags)) {
3622 size_t size = 0;
3623 munmap(s->data, s->size);
3624 s->should_munmap = 0;
3625 s->data = strbuf_detach(&buf, &size);
3626 s->size = size;
3627 s->should_free = 1;
3628 }
3629 }
3630 else {
3631 enum object_type type;
3632 if (size_only || (flags & CHECK_BINARY)) {
3633 type = oid_object_info(the_repository, &s->oid,
3634 &s->size);
3635 if (type < 0)
3636 die("unable to read %s",
3637 oid_to_hex(&s->oid));
3638 if (size_only)
3639 return 0;
3640 if (s->size > big_file_threshold && s->is_binary == -1) {
3641 s->is_binary = 1;
3642 return 0;
3643 }
3644 }
3645 s->data = read_object_file(&s->oid, &type, &s->size);
3646 if (!s->data)
3647 die("unable to read %s", oid_to_hex(&s->oid));
3648 s->should_free = 1;
3649 }
3650 return 0;
3651 }
3652
3653 void diff_free_filespec_blob(struct diff_filespec *s)
3654 {
3655 if (s->should_free)
3656 free(s->data);
3657 else if (s->should_munmap)
3658 munmap(s->data, s->size);
3659
3660 if (s->should_free || s->should_munmap) {
3661 s->should_free = s->should_munmap = 0;
3662 s->data = NULL;
3663 }
3664 }
3665
3666 void diff_free_filespec_data(struct diff_filespec *s)
3667 {
3668 diff_free_filespec_blob(s);
3669 FREE_AND_NULL(s->cnt_data);
3670 }
3671
3672 static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
3673 void *blob,
3674 unsigned long size,
3675 const struct object_id *oid,
3676 int mode)
3677 {
3678 struct strbuf buf = STRBUF_INIT;
3679 struct strbuf tempfile = STRBUF_INIT;
3680 char *path_dup = xstrdup(path);
3681 const char *base = basename(path_dup);
3682
3683 /* Generate "XXXXXX_basename.ext" */
3684 strbuf_addstr(&tempfile, "XXXXXX_");
3685 strbuf_addstr(&tempfile, base);
3686
3687 temp->tempfile = mks_tempfile_ts(tempfile.buf, strlen(base) + 1);
3688 if (!temp->tempfile)
3689 die_errno("unable to create temp-file");
3690 if (convert_to_working_tree(path,
3691 (const char *)blob, (size_t)size, &buf)) {
3692 blob = buf.buf;
3693 size = buf.len;
3694 }
3695 if (write_in_full(temp->tempfile->fd, blob, size) < 0 ||
3696 close_tempfile_gently(temp->tempfile))
3697 die_errno("unable to write temp-file");
3698 temp->name = get_tempfile_path(temp->tempfile);
3699 oid_to_hex_r(temp->hex, oid);
3700 xsnprintf(temp->mode, sizeof(temp->mode), "%06o", mode);
3701 strbuf_release(&buf);
3702 strbuf_release(&tempfile);
3703 free(path_dup);
3704 }
3705
3706 static struct diff_tempfile *prepare_temp_file(const char *name,
3707 struct diff_filespec *one)
3708 {
3709 struct diff_tempfile *temp = claim_diff_tempfile();
3710
3711 if (!DIFF_FILE_VALID(one)) {
3712 not_a_valid_file:
3713 /* A '-' entry produces this for file-2, and
3714 * a '+' entry produces this for file-1.
3715 */
3716 temp->name = "/dev/null";
3717 xsnprintf(temp->hex, sizeof(temp->hex), ".");
3718 xsnprintf(temp->mode, sizeof(temp->mode), ".");
3719 return temp;
3720 }
3721
3722 if (!S_ISGITLINK(one->mode) &&
3723 (!one->oid_valid ||
3724 reuse_worktree_file(name, &one->oid, 1))) {
3725 struct stat st;
3726 if (lstat(name, &st) < 0) {
3727 if (errno == ENOENT)
3728 goto not_a_valid_file;
3729 die_errno("stat(%s)", name);
3730 }
3731 if (S_ISLNK(st.st_mode)) {
3732 struct strbuf sb = STRBUF_INIT;
3733 if (strbuf_readlink(&sb, name, st.st_size) < 0)
3734 die_errno("readlink(%s)", name);
3735 pr