Code

diff --color-words -U0: fix the location of hunk headers
[git.git] / diff.c
1 /*
2  * Copyright (C) 2005 Junio C Hamano
3  */
4 #include "cache.h"
5 #include "quote.h"
6 #include "diff.h"
7 #include "diffcore.h"
8 #include "delta.h"
9 #include "xdiff-interface.h"
10 #include "color.h"
11 #include "attr.h"
12 #include "run-command.h"
13 #include "utf8.h"
14 #include "userdiff.h"
15 #include "sigchain.h"
17 #ifdef NO_FAST_WORKING_DIRECTORY
18 #define FAST_WORKING_DIRECTORY 0
19 #else
20 #define FAST_WORKING_DIRECTORY 1
21 #endif
23 static int diff_detect_rename_default;
24 static int diff_rename_limit_default = 200;
25 static int diff_suppress_blank_empty;
26 int diff_use_color_default = -1;
27 static const char *diff_word_regex_cfg;
28 static const char *external_diff_cmd_cfg;
29 int diff_auto_refresh_index = 1;
30 static int diff_mnemonic_prefix;
32 static char diff_colors[][COLOR_MAXLEN] = {
33         GIT_COLOR_RESET,
34         GIT_COLOR_NORMAL,       /* PLAIN */
35         GIT_COLOR_BOLD,         /* METAINFO */
36         GIT_COLOR_CYAN,         /* FRAGINFO */
37         GIT_COLOR_RED,          /* OLD */
38         GIT_COLOR_GREEN,        /* NEW */
39         GIT_COLOR_YELLOW,       /* COMMIT */
40         GIT_COLOR_BG_RED,       /* WHITESPACE */
41 };
43 static void diff_filespec_load_driver(struct diff_filespec *one);
44 static char *run_textconv(const char *, struct diff_filespec *, size_t *);
46 static int parse_diff_color_slot(const char *var, int ofs)
47 {
48         if (!strcasecmp(var+ofs, "plain"))
49                 return DIFF_PLAIN;
50         if (!strcasecmp(var+ofs, "meta"))
51                 return DIFF_METAINFO;
52         if (!strcasecmp(var+ofs, "frag"))
53                 return DIFF_FRAGINFO;
54         if (!strcasecmp(var+ofs, "old"))
55                 return DIFF_FILE_OLD;
56         if (!strcasecmp(var+ofs, "new"))
57                 return DIFF_FILE_NEW;
58         if (!strcasecmp(var+ofs, "commit"))
59                 return DIFF_COMMIT;
60         if (!strcasecmp(var+ofs, "whitespace"))
61                 return DIFF_WHITESPACE;
62         die("bad config variable '%s'", var);
63 }
65 static int git_config_rename(const char *var, const char *value)
66 {
67         if (!value)
68                 return DIFF_DETECT_RENAME;
69         if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
70                 return  DIFF_DETECT_COPY;
71         return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
72 }
74 /*
75  * These are to give UI layer defaults.
76  * The core-level commands such as git-diff-files should
77  * never be affected by the setting of diff.renames
78  * the user happens to have in the configuration file.
79  */
80 int git_diff_ui_config(const char *var, const char *value, void *cb)
81 {
82         if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
83                 diff_use_color_default = git_config_colorbool(var, value, -1);
84                 return 0;
85         }
86         if (!strcmp(var, "diff.renames")) {
87                 diff_detect_rename_default = git_config_rename(var, value);
88                 return 0;
89         }
90         if (!strcmp(var, "diff.autorefreshindex")) {
91                 diff_auto_refresh_index = git_config_bool(var, value);
92                 return 0;
93         }
94         if (!strcmp(var, "diff.mnemonicprefix")) {
95                 diff_mnemonic_prefix = git_config_bool(var, value);
96                 return 0;
97         }
98         if (!strcmp(var, "diff.external"))
99                 return git_config_string(&external_diff_cmd_cfg, var, value);
100         if (!strcmp(var, "diff.wordregex"))
101                 return git_config_string(&diff_word_regex_cfg, var, value);
103         return git_diff_basic_config(var, value, cb);
106 int git_diff_basic_config(const char *var, const char *value, void *cb)
108         if (!strcmp(var, "diff.renamelimit")) {
109                 diff_rename_limit_default = git_config_int(var, value);
110                 return 0;
111         }
113         switch (userdiff_config(var, value)) {
114                 case 0: break;
115                 case -1: return -1;
116                 default: return 0;
117         }
119         if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
120                 int slot = parse_diff_color_slot(var, 11);
121                 if (!value)
122                         return config_error_nonbool(var);
123                 color_parse(value, var, diff_colors[slot]);
124                 return 0;
125         }
127         /* like GNU diff's --suppress-blank-empty option  */
128         if (!strcmp(var, "diff.suppressblankempty") ||
129                         /* for backwards compatibility */
130                         !strcmp(var, "diff.suppress-blank-empty")) {
131                 diff_suppress_blank_empty = git_config_bool(var, value);
132                 return 0;
133         }
135         return git_color_default_config(var, value, cb);
138 static char *quote_two(const char *one, const char *two)
140         int need_one = quote_c_style(one, NULL, NULL, 1);
141         int need_two = quote_c_style(two, NULL, NULL, 1);
142         struct strbuf res = STRBUF_INIT;
144         if (need_one + need_two) {
145                 strbuf_addch(&res, '"');
146                 quote_c_style(one, &res, NULL, 1);
147                 quote_c_style(two, &res, NULL, 1);
148                 strbuf_addch(&res, '"');
149         } else {
150                 strbuf_addstr(&res, one);
151                 strbuf_addstr(&res, two);
152         }
153         return strbuf_detach(&res, NULL);
156 static const char *external_diff(void)
158         static const char *external_diff_cmd = NULL;
159         static int done_preparing = 0;
161         if (done_preparing)
162                 return external_diff_cmd;
163         external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
164         if (!external_diff_cmd)
165                 external_diff_cmd = external_diff_cmd_cfg;
166         done_preparing = 1;
167         return external_diff_cmd;
170 static struct diff_tempfile {
171         const char *name; /* filename external diff should read from */
172         char hex[41];
173         char mode[10];
174         char tmp_path[PATH_MAX];
175 } diff_temp[2];
177 static struct diff_tempfile *claim_diff_tempfile(void) {
178         int i;
179         for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
180                 if (!diff_temp[i].name)
181                         return diff_temp + i;
182         die("BUG: diff is failing to clean up its tempfiles");
185 static int remove_tempfile_installed;
187 static void remove_tempfile(void)
189         int i;
190         for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
191                 if (diff_temp[i].name == diff_temp[i].tmp_path)
192                         unlink_or_warn(diff_temp[i].name);
193                 diff_temp[i].name = NULL;
194         }
197 static void remove_tempfile_on_signal(int signo)
199         remove_tempfile();
200         sigchain_pop(signo);
201         raise(signo);
204 static int count_lines(const char *data, int size)
206         int count, ch, completely_empty = 1, nl_just_seen = 0;
207         count = 0;
208         while (0 < size--) {
209                 ch = *data++;
210                 if (ch == '\n') {
211                         count++;
212                         nl_just_seen = 1;
213                         completely_empty = 0;
214                 }
215                 else {
216                         nl_just_seen = 0;
217                         completely_empty = 0;
218                 }
219         }
220         if (completely_empty)
221                 return 0;
222         if (!nl_just_seen)
223                 count++; /* no trailing newline */
224         return count;
227 static void print_line_count(FILE *file, int count)
229         switch (count) {
230         case 0:
231                 fprintf(file, "0,0");
232                 break;
233         case 1:
234                 fprintf(file, "1");
235                 break;
236         default:
237                 fprintf(file, "1,%d", count);
238                 break;
239         }
242 static void copy_file_with_prefix(FILE *file,
243                                   int prefix, const char *data, int size,
244                                   const char *set, const char *reset)
246         int ch, nl_just_seen = 1;
247         while (0 < size--) {
248                 ch = *data++;
249                 if (nl_just_seen) {
250                         fputs(set, file);
251                         putc(prefix, file);
252                 }
253                 if (ch == '\n') {
254                         nl_just_seen = 1;
255                         fputs(reset, file);
256                 } else
257                         nl_just_seen = 0;
258                 putc(ch, file);
259         }
260         if (!nl_just_seen)
261                 fprintf(file, "%s\n\\ No newline at end of file\n", reset);
264 static void emit_rewrite_diff(const char *name_a,
265                               const char *name_b,
266                               struct diff_filespec *one,
267                               struct diff_filespec *two,
268                               const char *textconv_one,
269                               const char *textconv_two,
270                               struct diff_options *o)
272         int lc_a, lc_b;
273         int color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
274         const char *name_a_tab, *name_b_tab;
275         const char *metainfo = diff_get_color(color_diff, DIFF_METAINFO);
276         const char *fraginfo = diff_get_color(color_diff, DIFF_FRAGINFO);
277         const char *old = diff_get_color(color_diff, DIFF_FILE_OLD);
278         const char *new = diff_get_color(color_diff, DIFF_FILE_NEW);
279         const char *reset = diff_get_color(color_diff, DIFF_RESET);
280         static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
281         const char *a_prefix, *b_prefix;
282         const char *data_one, *data_two;
283         size_t size_one, size_two;
285         if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
286                 a_prefix = o->b_prefix;
287                 b_prefix = o->a_prefix;
288         } else {
289                 a_prefix = o->a_prefix;
290                 b_prefix = o->b_prefix;
291         }
293         name_a += (*name_a == '/');
294         name_b += (*name_b == '/');
295         name_a_tab = strchr(name_a, ' ') ? "\t" : "";
296         name_b_tab = strchr(name_b, ' ') ? "\t" : "";
298         strbuf_reset(&a_name);
299         strbuf_reset(&b_name);
300         quote_two_c_style(&a_name, a_prefix, name_a, 0);
301         quote_two_c_style(&b_name, b_prefix, name_b, 0);
303         diff_populate_filespec(one, 0);
304         diff_populate_filespec(two, 0);
305         if (textconv_one) {
306                 data_one = run_textconv(textconv_one, one, &size_one);
307                 if (!data_one)
308                         die("unable to read files to diff");
309         }
310         else {
311                 data_one = one->data;
312                 size_one = one->size;
313         }
314         if (textconv_two) {
315                 data_two = run_textconv(textconv_two, two, &size_two);
316                 if (!data_two)
317                         die("unable to read files to diff");
318         }
319         else {
320                 data_two = two->data;
321                 size_two = two->size;
322         }
324         lc_a = count_lines(data_one, size_one);
325         lc_b = count_lines(data_two, size_two);
326         fprintf(o->file,
327                 "%s--- %s%s%s\n%s+++ %s%s%s\n%s@@ -",
328                 metainfo, a_name.buf, name_a_tab, reset,
329                 metainfo, b_name.buf, name_b_tab, reset, fraginfo);
330         print_line_count(o->file, lc_a);
331         fprintf(o->file, " +");
332         print_line_count(o->file, lc_b);
333         fprintf(o->file, " @@%s\n", reset);
334         if (lc_a)
335                 copy_file_with_prefix(o->file, '-', data_one, size_one, old, reset);
336         if (lc_b)
337                 copy_file_with_prefix(o->file, '+', data_two, size_two, new, reset);
340 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
342         if (!DIFF_FILE_VALID(one)) {
343                 mf->ptr = (char *)""; /* does not matter */
344                 mf->size = 0;
345                 return 0;
346         }
347         else if (diff_populate_filespec(one, 0))
348                 return -1;
350         mf->ptr = one->data;
351         mf->size = one->size;
352         return 0;
355 struct diff_words_buffer {
356         mmfile_t text;
357         long alloc;
358         struct diff_words_orig {
359                 const char *begin, *end;
360         } *orig;
361         int orig_nr, orig_alloc;
362 };
364 static void diff_words_append(char *line, unsigned long len,
365                 struct diff_words_buffer *buffer)
367         ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
368         line++;
369         len--;
370         memcpy(buffer->text.ptr + buffer->text.size, line, len);
371         buffer->text.size += len;
372         buffer->text.ptr[buffer->text.size] = '\0';
375 struct diff_words_data {
376         struct diff_words_buffer minus, plus;
377         const char *current_plus;
378         FILE *file;
379         regex_t *word_regex;
380 };
382 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
384         struct diff_words_data *diff_words = priv;
385         int minus_first, minus_len, plus_first, plus_len;
386         const char *minus_begin, *minus_end, *plus_begin, *plus_end;
388         if (line[0] != '@' || parse_hunk_header(line, len,
389                         &minus_first, &minus_len, &plus_first, &plus_len))
390                 return;
392         /* POSIX requires that first be decremented by one if len == 0... */
393         if (minus_len) {
394                 minus_begin = diff_words->minus.orig[minus_first].begin;
395                 minus_end =
396                         diff_words->minus.orig[minus_first + minus_len - 1].end;
397         } else
398                 minus_begin = minus_end =
399                         diff_words->minus.orig[minus_first].end;
401         if (plus_len) {
402                 plus_begin = diff_words->plus.orig[plus_first].begin;
403                 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
404         } else
405                 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
407         if (diff_words->current_plus != plus_begin)
408                 fwrite(diff_words->current_plus,
409                                 plus_begin - diff_words->current_plus, 1,
410                                 diff_words->file);
411         if (minus_begin != minus_end)
412                 color_fwrite_lines(diff_words->file,
413                                 diff_get_color(1, DIFF_FILE_OLD),
414                                 minus_end - minus_begin, minus_begin);
415         if (plus_begin != plus_end)
416                 color_fwrite_lines(diff_words->file,
417                                 diff_get_color(1, DIFF_FILE_NEW),
418                                 plus_end - plus_begin, plus_begin);
420         diff_words->current_plus = plus_end;
423 /* This function starts looking at *begin, and returns 0 iff a word was found. */
424 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
425                 int *begin, int *end)
427         if (word_regex && *begin < buffer->size) {
428                 regmatch_t match[1];
429                 if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
430                         char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
431                                         '\n', match[0].rm_eo - match[0].rm_so);
432                         *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
433                         *begin += match[0].rm_so;
434                         return *begin >= *end;
435                 }
436                 return -1;
437         }
439         /* find the next word */
440         while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
441                 (*begin)++;
442         if (*begin >= buffer->size)
443                 return -1;
445         /* find the end of the word */
446         *end = *begin + 1;
447         while (*end < buffer->size && !isspace(buffer->ptr[*end]))
448                 (*end)++;
450         return 0;
453 /*
454  * This function splits the words in buffer->text, stores the list with
455  * newline separator into out, and saves the offsets of the original words
456  * in buffer->orig.
457  */
458 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
459                 regex_t *word_regex)
461         int i, j;
462         long alloc = 0;
464         out->size = 0;
465         out->ptr = NULL;
467         /* fake an empty "0th" word */
468         ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
469         buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
470         buffer->orig_nr = 1;
472         for (i = 0; i < buffer->text.size; i++) {
473                 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
474                         return;
476                 /* store original boundaries */
477                 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
478                                 buffer->orig_alloc);
479                 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
480                 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
481                 buffer->orig_nr++;
483                 /* store one word */
484                 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
485                 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
486                 out->ptr[out->size + j - i] = '\n';
487                 out->size += j - i + 1;
489                 i = j - 1;
490         }
493 /* this executes the word diff on the accumulated buffers */
494 static void diff_words_show(struct diff_words_data *diff_words)
496         xpparam_t xpp;
497         xdemitconf_t xecfg;
498         xdemitcb_t ecb;
499         mmfile_t minus, plus;
501         /* special case: only removal */
502         if (!diff_words->plus.text.size) {
503                 color_fwrite_lines(diff_words->file,
504                         diff_get_color(1, DIFF_FILE_OLD),
505                         diff_words->minus.text.size, diff_words->minus.text.ptr);
506                 diff_words->minus.text.size = 0;
507                 return;
508         }
510         diff_words->current_plus = diff_words->plus.text.ptr;
512         memset(&xpp, 0, sizeof(xpp));
513         memset(&xecfg, 0, sizeof(xecfg));
514         diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
515         diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
516         xpp.flags = XDF_NEED_MINIMAL;
517         /* as only the hunk header will be parsed, we need a 0-context */
518         xecfg.ctxlen = 0;
519         xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
520                       &xpp, &xecfg, &ecb);
521         free(minus.ptr);
522         free(plus.ptr);
523         if (diff_words->current_plus != diff_words->plus.text.ptr +
524                         diff_words->plus.text.size)
525                 fwrite(diff_words->current_plus,
526                         diff_words->plus.text.ptr + diff_words->plus.text.size
527                         - diff_words->current_plus, 1,
528                         diff_words->file);
529         diff_words->minus.text.size = diff_words->plus.text.size = 0;
532 typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
534 struct emit_callback {
535         int nparents, color_diff;
536         unsigned ws_rule;
537         sane_truncate_fn truncate;
538         const char **label_path;
539         struct diff_words_data *diff_words;
540         int *found_changesp;
541         FILE *file;
542 };
544 static void free_diff_words_data(struct emit_callback *ecbdata)
546         if (ecbdata->diff_words) {
547                 /* flush buffers */
548                 if (ecbdata->diff_words->minus.text.size ||
549                                 ecbdata->diff_words->plus.text.size)
550                         diff_words_show(ecbdata->diff_words);
552                 free (ecbdata->diff_words->minus.text.ptr);
553                 free (ecbdata->diff_words->minus.orig);
554                 free (ecbdata->diff_words->plus.text.ptr);
555                 free (ecbdata->diff_words->plus.orig);
556                 free(ecbdata->diff_words->word_regex);
557                 free(ecbdata->diff_words);
558                 ecbdata->diff_words = NULL;
559         }
562 const char *diff_get_color(int diff_use_color, enum color_diff ix)
564         if (diff_use_color)
565                 return diff_colors[ix];
566         return "";
569 static void emit_line(FILE *file, const char *set, const char *reset, const char *line, int len)
571         int has_trailing_newline, has_trailing_carriage_return;
573         has_trailing_newline = (len > 0 && line[len-1] == '\n');
574         if (has_trailing_newline)
575                 len--;
576         has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
577         if (has_trailing_carriage_return)
578                 len--;
580         fputs(set, file);
581         fwrite(line, len, 1, file);
582         fputs(reset, file);
583         if (has_trailing_carriage_return)
584                 fputc('\r', file);
585         if (has_trailing_newline)
586                 fputc('\n', file);
589 static void emit_add_line(const char *reset, struct emit_callback *ecbdata, const char *line, int len)
591         const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
592         const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
594         if (!*ws)
595                 emit_line(ecbdata->file, set, reset, line, len);
596         else {
597                 /* Emit just the prefix, then the rest. */
598                 emit_line(ecbdata->file, set, reset, line, ecbdata->nparents);
599                 ws_check_emit(line + ecbdata->nparents,
600                               len - ecbdata->nparents, ecbdata->ws_rule,
601                               ecbdata->file, set, reset, ws);
602         }
605 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
607         const char *cp;
608         unsigned long allot;
609         size_t l = len;
611         if (ecb->truncate)
612                 return ecb->truncate(line, len);
613         cp = line;
614         allot = l;
615         while (0 < l) {
616                 (void) utf8_width(&cp, &l);
617                 if (!cp)
618                         break; /* truncated in the middle? */
619         }
620         return allot - l;
623 static void fn_out_consume(void *priv, char *line, unsigned long len)
625         int i;
626         int color;
627         struct emit_callback *ecbdata = priv;
628         const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
629         const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
630         const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
632         *(ecbdata->found_changesp) = 1;
634         if (ecbdata->label_path[0]) {
635                 const char *name_a_tab, *name_b_tab;
637                 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
638                 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
640                 fprintf(ecbdata->file, "%s--- %s%s%s\n",
641                         meta, ecbdata->label_path[0], reset, name_a_tab);
642                 fprintf(ecbdata->file, "%s+++ %s%s%s\n",
643                         meta, ecbdata->label_path[1], reset, name_b_tab);
644                 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
645         }
647         if (diff_suppress_blank_empty
648             && len == 2 && line[0] == ' ' && line[1] == '\n') {
649                 line[0] = '\n';
650                 len = 1;
651         }
653         /* This is not really necessary for now because
654          * this codepath only deals with two-way diffs.
655          */
656         for (i = 0; i < len && line[i] == '@'; i++)
657                 ;
658         if (2 <= i && i < len && line[i] == ' ') {
659                 /* flush --color-words even for --unified=0 */
660                 if (ecbdata->diff_words &&
661                     (ecbdata->diff_words->minus.text.size ||
662                      ecbdata->diff_words->plus.text.size))
663                         diff_words_show(ecbdata->diff_words);
665                 ecbdata->nparents = i - 1;
666                 len = sane_truncate_line(ecbdata, line, len);
667                 emit_line(ecbdata->file,
668                           diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO),
669                           reset, line, len);
670                 if (line[len-1] != '\n')
671                         putc('\n', ecbdata->file);
672                 return;
673         }
675         if (len < ecbdata->nparents) {
676                 emit_line(ecbdata->file, reset, reset, line, len);
677                 return;
678         }
680         color = DIFF_PLAIN;
681         if (ecbdata->diff_words && ecbdata->nparents != 1)
682                 /* fall back to normal diff */
683                 free_diff_words_data(ecbdata);
684         if (ecbdata->diff_words) {
685                 if (line[0] == '-') {
686                         diff_words_append(line, len,
687                                           &ecbdata->diff_words->minus);
688                         return;
689                 } else if (line[0] == '+') {
690                         diff_words_append(line, len,
691                                           &ecbdata->diff_words->plus);
692                         return;
693                 }
694                 if (ecbdata->diff_words->minus.text.size ||
695                     ecbdata->diff_words->plus.text.size)
696                         diff_words_show(ecbdata->diff_words);
697                 line++;
698                 len--;
699                 emit_line(ecbdata->file, plain, reset, line, len);
700                 return;
701         }
702         for (i = 0; i < ecbdata->nparents && len; i++) {
703                 if (line[i] == '-')
704                         color = DIFF_FILE_OLD;
705                 else if (line[i] == '+')
706                         color = DIFF_FILE_NEW;
707         }
709         if (color != DIFF_FILE_NEW) {
710                 emit_line(ecbdata->file,
711                           diff_get_color(ecbdata->color_diff, color),
712                           reset, line, len);
713                 return;
714         }
715         emit_add_line(reset, ecbdata, line, len);
718 static char *pprint_rename(const char *a, const char *b)
720         const char *old = a;
721         const char *new = b;
722         struct strbuf name = STRBUF_INIT;
723         int pfx_length, sfx_length;
724         int len_a = strlen(a);
725         int len_b = strlen(b);
726         int a_midlen, b_midlen;
727         int qlen_a = quote_c_style(a, NULL, NULL, 0);
728         int qlen_b = quote_c_style(b, NULL, NULL, 0);
730         if (qlen_a || qlen_b) {
731                 quote_c_style(a, &name, NULL, 0);
732                 strbuf_addstr(&name, " => ");
733                 quote_c_style(b, &name, NULL, 0);
734                 return strbuf_detach(&name, NULL);
735         }
737         /* Find common prefix */
738         pfx_length = 0;
739         while (*old && *new && *old == *new) {
740                 if (*old == '/')
741                         pfx_length = old - a + 1;
742                 old++;
743                 new++;
744         }
746         /* Find common suffix */
747         old = a + len_a;
748         new = b + len_b;
749         sfx_length = 0;
750         while (a <= old && b <= new && *old == *new) {
751                 if (*old == '/')
752                         sfx_length = len_a - (old - a);
753                 old--;
754                 new--;
755         }
757         /*
758          * pfx{mid-a => mid-b}sfx
759          * {pfx-a => pfx-b}sfx
760          * pfx{sfx-a => sfx-b}
761          * name-a => name-b
762          */
763         a_midlen = len_a - pfx_length - sfx_length;
764         b_midlen = len_b - pfx_length - sfx_length;
765         if (a_midlen < 0)
766                 a_midlen = 0;
767         if (b_midlen < 0)
768                 b_midlen = 0;
770         strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
771         if (pfx_length + sfx_length) {
772                 strbuf_add(&name, a, pfx_length);
773                 strbuf_addch(&name, '{');
774         }
775         strbuf_add(&name, a + pfx_length, a_midlen);
776         strbuf_addstr(&name, " => ");
777         strbuf_add(&name, b + pfx_length, b_midlen);
778         if (pfx_length + sfx_length) {
779                 strbuf_addch(&name, '}');
780                 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
781         }
782         return strbuf_detach(&name, NULL);
785 struct diffstat_t {
786         int nr;
787         int alloc;
788         struct diffstat_file {
789                 char *from_name;
790                 char *name;
791                 char *print_name;
792                 unsigned is_unmerged:1;
793                 unsigned is_binary:1;
794                 unsigned is_renamed:1;
795                 unsigned int added, deleted;
796         } **files;
797 };
799 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
800                                           const char *name_a,
801                                           const char *name_b)
803         struct diffstat_file *x;
804         x = xcalloc(sizeof (*x), 1);
805         if (diffstat->nr == diffstat->alloc) {
806                 diffstat->alloc = alloc_nr(diffstat->alloc);
807                 diffstat->files = xrealloc(diffstat->files,
808                                 diffstat->alloc * sizeof(x));
809         }
810         diffstat->files[diffstat->nr++] = x;
811         if (name_b) {
812                 x->from_name = xstrdup(name_a);
813                 x->name = xstrdup(name_b);
814                 x->is_renamed = 1;
815         }
816         else {
817                 x->from_name = NULL;
818                 x->name = xstrdup(name_a);
819         }
820         return x;
823 static void diffstat_consume(void *priv, char *line, unsigned long len)
825         struct diffstat_t *diffstat = priv;
826         struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
828         if (line[0] == '+')
829                 x->added++;
830         else if (line[0] == '-')
831                 x->deleted++;
834 const char mime_boundary_leader[] = "------------";
836 static int scale_linear(int it, int width, int max_change)
838         /*
839          * make sure that at least one '-' is printed if there were deletions,
840          * and likewise for '+'.
841          */
842         if (max_change < 2)
843                 return it;
844         return ((it - 1) * (width - 1) + max_change - 1) / (max_change - 1);
847 static void show_name(FILE *file,
848                       const char *prefix, const char *name, int len)
850         fprintf(file, " %s%-*s |", prefix, len, name);
853 static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
855         if (cnt <= 0)
856                 return;
857         fprintf(file, "%s", set);
858         while (cnt--)
859                 putc(ch, file);
860         fprintf(file, "%s", reset);
863 static void fill_print_name(struct diffstat_file *file)
865         char *pname;
867         if (file->print_name)
868                 return;
870         if (!file->is_renamed) {
871                 struct strbuf buf = STRBUF_INIT;
872                 if (quote_c_style(file->name, &buf, NULL, 0)) {
873                         pname = strbuf_detach(&buf, NULL);
874                 } else {
875                         pname = file->name;
876                         strbuf_release(&buf);
877                 }
878         } else {
879                 pname = pprint_rename(file->from_name, file->name);
880         }
881         file->print_name = pname;
884 static void show_stats(struct diffstat_t *data, struct diff_options *options)
886         int i, len, add, del, adds = 0, dels = 0;
887         int max_change = 0, max_len = 0;
888         int total_files = data->nr;
889         int width, name_width;
890         const char *reset, *set, *add_c, *del_c;
892         if (data->nr == 0)
893                 return;
895         width = options->stat_width ? options->stat_width : 80;
896         name_width = options->stat_name_width ? options->stat_name_width : 50;
898         /* Sanity: give at least 5 columns to the graph,
899          * but leave at least 10 columns for the name.
900          */
901         if (width < 25)
902                 width = 25;
903         if (name_width < 10)
904                 name_width = 10;
905         else if (width < name_width + 15)
906                 name_width = width - 15;
908         /* Find the longest filename and max number of changes */
909         reset = diff_get_color_opt(options, DIFF_RESET);
910         set   = diff_get_color_opt(options, DIFF_PLAIN);
911         add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
912         del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
914         for (i = 0; i < data->nr; i++) {
915                 struct diffstat_file *file = data->files[i];
916                 int change = file->added + file->deleted;
917                 fill_print_name(file);
918                 len = strlen(file->print_name);
919                 if (max_len < len)
920                         max_len = len;
922                 if (file->is_binary || file->is_unmerged)
923                         continue;
924                 if (max_change < change)
925                         max_change = change;
926         }
928         /* Compute the width of the graph part;
929          * 10 is for one blank at the beginning of the line plus
930          * " | count " between the name and the graph.
931          *
932          * From here on, name_width is the width of the name area,
933          * and width is the width of the graph area.
934          */
935         name_width = (name_width < max_len) ? name_width : max_len;
936         if (width < (name_width + 10) + max_change)
937                 width = width - (name_width + 10);
938         else
939                 width = max_change;
941         for (i = 0; i < data->nr; i++) {
942                 const char *prefix = "";
943                 char *name = data->files[i]->print_name;
944                 int added = data->files[i]->added;
945                 int deleted = data->files[i]->deleted;
946                 int name_len;
948                 /*
949                  * "scale" the filename
950                  */
951                 len = name_width;
952                 name_len = strlen(name);
953                 if (name_width < name_len) {
954                         char *slash;
955                         prefix = "...";
956                         len -= 3;
957                         name += name_len - len;
958                         slash = strchr(name, '/');
959                         if (slash)
960                                 name = slash;
961                 }
963                 if (data->files[i]->is_binary) {
964                         show_name(options->file, prefix, name, len);
965                         fprintf(options->file, "  Bin ");
966                         fprintf(options->file, "%s%d%s", del_c, deleted, reset);
967                         fprintf(options->file, " -> ");
968                         fprintf(options->file, "%s%d%s", add_c, added, reset);
969                         fprintf(options->file, " bytes");
970                         fprintf(options->file, "\n");
971                         continue;
972                 }
973                 else if (data->files[i]->is_unmerged) {
974                         show_name(options->file, prefix, name, len);
975                         fprintf(options->file, "  Unmerged\n");
976                         continue;
977                 }
978                 else if (!data->files[i]->is_renamed &&
979                          (added + deleted == 0)) {
980                         total_files--;
981                         continue;
982                 }
984                 /*
985                  * scale the add/delete
986                  */
987                 add = added;
988                 del = deleted;
989                 adds += add;
990                 dels += del;
992                 if (width <= max_change) {
993                         add = scale_linear(add, width, max_change);
994                         del = scale_linear(del, width, max_change);
995                 }
996                 show_name(options->file, prefix, name, len);
997                 fprintf(options->file, "%5d%s", added + deleted,
998                                 added + deleted ? " " : "");
999                 show_graph(options->file, '+', add, add_c, reset);
1000                 show_graph(options->file, '-', del, del_c, reset);
1001                 fprintf(options->file, "\n");
1002         }
1003         fprintf(options->file,
1004                " %d files changed, %d insertions(+), %d deletions(-)\n",
1005                total_files, adds, dels);
1008 static void show_shortstats(struct diffstat_t* data, struct diff_options *options)
1010         int i, adds = 0, dels = 0, total_files = data->nr;
1012         if (data->nr == 0)
1013                 return;
1015         for (i = 0; i < data->nr; i++) {
1016                 if (!data->files[i]->is_binary &&
1017                     !data->files[i]->is_unmerged) {
1018                         int added = data->files[i]->added;
1019                         int deleted= data->files[i]->deleted;
1020                         if (!data->files[i]->is_renamed &&
1021                             (added + deleted == 0)) {
1022                                 total_files--;
1023                         } else {
1024                                 adds += added;
1025                                 dels += deleted;
1026                         }
1027                 }
1028         }
1029         fprintf(options->file, " %d files changed, %d insertions(+), %d deletions(-)\n",
1030                total_files, adds, dels);
1033 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1035         int i;
1037         if (data->nr == 0)
1038                 return;
1040         for (i = 0; i < data->nr; i++) {
1041                 struct diffstat_file *file = data->files[i];
1043                 if (file->is_binary)
1044                         fprintf(options->file, "-\t-\t");
1045                 else
1046                         fprintf(options->file,
1047                                 "%d\t%d\t", file->added, file->deleted);
1048                 if (options->line_termination) {
1049                         fill_print_name(file);
1050                         if (!file->is_renamed)
1051                                 write_name_quoted(file->name, options->file,
1052                                                   options->line_termination);
1053                         else {
1054                                 fputs(file->print_name, options->file);
1055                                 putc(options->line_termination, options->file);
1056                         }
1057                 } else {
1058                         if (file->is_renamed) {
1059                                 putc('\0', options->file);
1060                                 write_name_quoted(file->from_name, options->file, '\0');
1061                         }
1062                         write_name_quoted(file->name, options->file, '\0');
1063                 }
1064         }
1067 struct dirstat_file {
1068         const char *name;
1069         unsigned long changed;
1070 };
1072 struct dirstat_dir {
1073         struct dirstat_file *files;
1074         int alloc, nr, percent, cumulative;
1075 };
1077 static long gather_dirstat(FILE *file, struct dirstat_dir *dir, unsigned long changed, const char *base, int baselen)
1079         unsigned long this_dir = 0;
1080         unsigned int sources = 0;
1082         while (dir->nr) {
1083                 struct dirstat_file *f = dir->files;
1084                 int namelen = strlen(f->name);
1085                 unsigned long this;
1086                 char *slash;
1088                 if (namelen < baselen)
1089                         break;
1090                 if (memcmp(f->name, base, baselen))
1091                         break;
1092                 slash = strchr(f->name + baselen, '/');
1093                 if (slash) {
1094                         int newbaselen = slash + 1 - f->name;
1095                         this = gather_dirstat(file, dir, changed, f->name, newbaselen);
1096                         sources++;
1097                 } else {
1098                         this = f->changed;
1099                         dir->files++;
1100                         dir->nr--;
1101                         sources += 2;
1102                 }
1103                 this_dir += this;
1104         }
1106         /*
1107          * We don't report dirstat's for
1108          *  - the top level
1109          *  - or cases where everything came from a single directory
1110          *    under this directory (sources == 1).
1111          */
1112         if (baselen && sources != 1) {
1113                 int permille = this_dir * 1000 / changed;
1114                 if (permille) {
1115                         int percent = permille / 10;
1116                         if (percent >= dir->percent) {
1117                                 fprintf(file, "%4d.%01d%% %.*s\n", percent, permille % 10, baselen, base);
1118                                 if (!dir->cumulative)
1119                                         return 0;
1120                         }
1121                 }
1122         }
1123         return this_dir;
1126 static int dirstat_compare(const void *_a, const void *_b)
1128         const struct dirstat_file *a = _a;
1129         const struct dirstat_file *b = _b;
1130         return strcmp(a->name, b->name);
1133 static void show_dirstat(struct diff_options *options)
1135         int i;
1136         unsigned long changed;
1137         struct dirstat_dir dir;
1138         struct diff_queue_struct *q = &diff_queued_diff;
1140         dir.files = NULL;
1141         dir.alloc = 0;
1142         dir.nr = 0;
1143         dir.percent = options->dirstat_percent;
1144         dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1146         changed = 0;
1147         for (i = 0; i < q->nr; i++) {
1148                 struct diff_filepair *p = q->queue[i];
1149                 const char *name;
1150                 unsigned long copied, added, damage;
1152                 name = p->one->path ? p->one->path : p->two->path;
1154                 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1155                         diff_populate_filespec(p->one, 0);
1156                         diff_populate_filespec(p->two, 0);
1157                         diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1158                                                &copied, &added);
1159                         diff_free_filespec_data(p->one);
1160                         diff_free_filespec_data(p->two);
1161                 } else if (DIFF_FILE_VALID(p->one)) {
1162                         diff_populate_filespec(p->one, 1);
1163                         copied = added = 0;
1164                         diff_free_filespec_data(p->one);
1165                 } else if (DIFF_FILE_VALID(p->two)) {
1166                         diff_populate_filespec(p->two, 1);
1167                         copied = 0;
1168                         added = p->two->size;
1169                         diff_free_filespec_data(p->two);
1170                 } else
1171                         continue;
1173                 /*
1174                  * Original minus copied is the removed material,
1175                  * added is the new material.  They are both damages
1176                  * made to the preimage. In --dirstat-by-file mode, count
1177                  * damaged files, not damaged lines. This is done by
1178                  * counting only a single damaged line per file.
1179                  */
1180                 damage = (p->one->size - copied) + added;
1181                 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE) && damage > 0)
1182                         damage = 1;
1184                 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1185                 dir.files[dir.nr].name = name;
1186                 dir.files[dir.nr].changed = damage;
1187                 changed += damage;
1188                 dir.nr++;
1189         }
1191         /* This can happen even with many files, if everything was renames */
1192         if (!changed)
1193                 return;
1195         /* Show all directories with more than x% of the changes */
1196         qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1197         gather_dirstat(options->file, &dir, changed, "", 0);
1200 static void free_diffstat_info(struct diffstat_t *diffstat)
1202         int i;
1203         for (i = 0; i < diffstat->nr; i++) {
1204                 struct diffstat_file *f = diffstat->files[i];
1205                 if (f->name != f->print_name)
1206                         free(f->print_name);
1207                 free(f->name);
1208                 free(f->from_name);
1209                 free(f);
1210         }
1211         free(diffstat->files);
1214 struct checkdiff_t {
1215         const char *filename;
1216         int lineno;
1217         struct diff_options *o;
1218         unsigned ws_rule;
1219         unsigned status;
1220         int trailing_blanks_start;
1221 };
1223 static int is_conflict_marker(const char *line, unsigned long len)
1225         char firstchar;
1226         int cnt;
1228         if (len < 8)
1229                 return 0;
1230         firstchar = line[0];
1231         switch (firstchar) {
1232         case '=': case '>': case '<':
1233                 break;
1234         default:
1235                 return 0;
1236         }
1237         for (cnt = 1; cnt < 7; cnt++)
1238                 if (line[cnt] != firstchar)
1239                         return 0;
1240         /* line[0] thru line[6] are same as firstchar */
1241         if (firstchar == '=') {
1242                 /* divider between ours and theirs? */
1243                 if (len != 8 || line[7] != '\n')
1244                         return 0;
1245         } else if (len < 8 || !isspace(line[7])) {
1246                 /* not divider before ours nor after theirs */
1247                 return 0;
1248         }
1249         return 1;
1252 static void checkdiff_consume(void *priv, char *line, unsigned long len)
1254         struct checkdiff_t *data = priv;
1255         int color_diff = DIFF_OPT_TST(data->o, COLOR_DIFF);
1256         const char *ws = diff_get_color(color_diff, DIFF_WHITESPACE);
1257         const char *reset = diff_get_color(color_diff, DIFF_RESET);
1258         const char *set = diff_get_color(color_diff, DIFF_FILE_NEW);
1259         char *err;
1261         if (line[0] == '+') {
1262                 unsigned bad;
1263                 data->lineno++;
1264                 if (!ws_blank_line(line + 1, len - 1, data->ws_rule))
1265                         data->trailing_blanks_start = 0;
1266                 else if (!data->trailing_blanks_start)
1267                         data->trailing_blanks_start = data->lineno;
1268                 if (is_conflict_marker(line + 1, len - 1)) {
1269                         data->status |= 1;
1270                         fprintf(data->o->file,
1271                                 "%s:%d: leftover conflict marker\n",
1272                                 data->filename, data->lineno);
1273                 }
1274                 bad = ws_check(line + 1, len - 1, data->ws_rule);
1275                 if (!bad)
1276                         return;
1277                 data->status |= bad;
1278                 err = whitespace_error_string(bad);
1279                 fprintf(data->o->file, "%s:%d: %s.\n",
1280                         data->filename, data->lineno, err);
1281                 free(err);
1282                 emit_line(data->o->file, set, reset, line, 1);
1283                 ws_check_emit(line + 1, len - 1, data->ws_rule,
1284                               data->o->file, set, reset, ws);
1285         } else if (line[0] == ' ') {
1286                 data->lineno++;
1287                 data->trailing_blanks_start = 0;
1288         } else if (line[0] == '@') {
1289                 char *plus = strchr(line, '+');
1290                 if (plus)
1291                         data->lineno = strtol(plus, NULL, 10) - 1;
1292                 else
1293                         die("invalid diff");
1294                 data->trailing_blanks_start = 0;
1295         }
1298 static unsigned char *deflate_it(char *data,
1299                                  unsigned long size,
1300                                  unsigned long *result_size)
1302         int bound;
1303         unsigned char *deflated;
1304         z_stream stream;
1306         memset(&stream, 0, sizeof(stream));
1307         deflateInit(&stream, zlib_compression_level);
1308         bound = deflateBound(&stream, size);
1309         deflated = xmalloc(bound);
1310         stream.next_out = deflated;
1311         stream.avail_out = bound;
1313         stream.next_in = (unsigned char *)data;
1314         stream.avail_in = size;
1315         while (deflate(&stream, Z_FINISH) == Z_OK)
1316                 ; /* nothing */
1317         deflateEnd(&stream);
1318         *result_size = stream.total_out;
1319         return deflated;
1322 static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two)
1324         void *cp;
1325         void *delta;
1326         void *deflated;
1327         void *data;
1328         unsigned long orig_size;
1329         unsigned long delta_size;
1330         unsigned long deflate_size;
1331         unsigned long data_size;
1333         /* We could do deflated delta, or we could do just deflated two,
1334          * whichever is smaller.
1335          */
1336         delta = NULL;
1337         deflated = deflate_it(two->ptr, two->size, &deflate_size);
1338         if (one->size && two->size) {
1339                 delta = diff_delta(one->ptr, one->size,
1340                                    two->ptr, two->size,
1341                                    &delta_size, deflate_size);
1342                 if (delta) {
1343                         void *to_free = delta;
1344                         orig_size = delta_size;
1345                         delta = deflate_it(delta, delta_size, &delta_size);
1346                         free(to_free);
1347                 }
1348         }
1350         if (delta && delta_size < deflate_size) {
1351                 fprintf(file, "delta %lu\n", orig_size);
1352                 free(deflated);
1353                 data = delta;
1354                 data_size = delta_size;
1355         }
1356         else {
1357                 fprintf(file, "literal %lu\n", two->size);
1358                 free(delta);
1359                 data = deflated;
1360                 data_size = deflate_size;
1361         }
1363         /* emit data encoded in base85 */
1364         cp = data;
1365         while (data_size) {
1366                 int bytes = (52 < data_size) ? 52 : data_size;
1367                 char line[70];
1368                 data_size -= bytes;
1369                 if (bytes <= 26)
1370                         line[0] = bytes + 'A' - 1;
1371                 else
1372                         line[0] = bytes - 26 + 'a' - 1;
1373                 encode_85(line + 1, cp, bytes);
1374                 cp = (char *) cp + bytes;
1375                 fputs(line, file);
1376                 fputc('\n', file);
1377         }
1378         fprintf(file, "\n");
1379         free(data);
1382 static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two)
1384         fprintf(file, "GIT binary patch\n");
1385         emit_binary_diff_body(file, one, two);
1386         emit_binary_diff_body(file, two, one);
1389 static void diff_filespec_load_driver(struct diff_filespec *one)
1391         if (!one->driver)
1392                 one->driver = userdiff_find_by_path(one->path);
1393         if (!one->driver)
1394                 one->driver = userdiff_find_by_name("default");
1397 int diff_filespec_is_binary(struct diff_filespec *one)
1399         if (one->is_binary == -1) {
1400                 diff_filespec_load_driver(one);
1401                 if (one->driver->binary != -1)
1402                         one->is_binary = one->driver->binary;
1403                 else {
1404                         if (!one->data && DIFF_FILE_VALID(one))
1405                                 diff_populate_filespec(one, 0);
1406                         if (one->data)
1407                                 one->is_binary = buffer_is_binary(one->data,
1408                                                 one->size);
1409                         if (one->is_binary == -1)
1410                                 one->is_binary = 0;
1411                 }
1412         }
1413         return one->is_binary;
1416 static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
1418         diff_filespec_load_driver(one);
1419         return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
1422 static const char *userdiff_word_regex(struct diff_filespec *one)
1424         diff_filespec_load_driver(one);
1425         return one->driver->word_regex;
1428 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
1430         if (!options->a_prefix)
1431                 options->a_prefix = a;
1432         if (!options->b_prefix)
1433                 options->b_prefix = b;
1436 static const char *get_textconv(struct diff_filespec *one)
1438         if (!DIFF_FILE_VALID(one))
1439                 return NULL;
1440         if (!S_ISREG(one->mode))
1441                 return NULL;
1442         diff_filespec_load_driver(one);
1443         return one->driver->textconv;
1446 static void builtin_diff(const char *name_a,
1447                          const char *name_b,
1448                          struct diff_filespec *one,
1449                          struct diff_filespec *two,
1450                          const char *xfrm_msg,
1451                          struct diff_options *o,
1452                          int complete_rewrite)
1454         mmfile_t mf1, mf2;
1455         const char *lbl[2];
1456         char *a_one, *b_two;
1457         const char *set = diff_get_color_opt(o, DIFF_METAINFO);
1458         const char *reset = diff_get_color_opt(o, DIFF_RESET);
1459         const char *a_prefix, *b_prefix;
1460         const char *textconv_one = NULL, *textconv_two = NULL;
1462         if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
1463                 textconv_one = get_textconv(one);
1464                 textconv_two = get_textconv(two);
1465         }
1467         diff_set_mnemonic_prefix(o, "a/", "b/");
1468         if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
1469                 a_prefix = o->b_prefix;
1470                 b_prefix = o->a_prefix;
1471         } else {
1472                 a_prefix = o->a_prefix;
1473                 b_prefix = o->b_prefix;
1474         }
1476         /* Never use a non-valid filename anywhere if at all possible */
1477         name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
1478         name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
1480         a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
1481         b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
1482         lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
1483         lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
1484         fprintf(o->file, "%sdiff --git %s %s%s\n", set, a_one, b_two, reset);
1485         if (lbl[0][0] == '/') {
1486                 /* /dev/null */
1487                 fprintf(o->file, "%snew file mode %06o%s\n", set, two->mode, reset);
1488                 if (xfrm_msg && xfrm_msg[0])
1489                         fprintf(o->file, "%s%s%s\n", set, xfrm_msg, reset);
1490         }
1491         else if (lbl[1][0] == '/') {
1492                 fprintf(o->file, "%sdeleted file mode %06o%s\n", set, one->mode, reset);
1493                 if (xfrm_msg && xfrm_msg[0])
1494                         fprintf(o->file, "%s%s%s\n", set, xfrm_msg, reset);
1495         }
1496         else {
1497                 if (one->mode != two->mode) {
1498                         fprintf(o->file, "%sold mode %06o%s\n", set, one->mode, reset);
1499                         fprintf(o->file, "%snew mode %06o%s\n", set, two->mode, reset);
1500                 }
1501                 if (xfrm_msg && xfrm_msg[0])
1502                         fprintf(o->file, "%s%s%s\n", set, xfrm_msg, reset);
1503                 /*
1504                  * we do not run diff between different kind
1505                  * of objects.
1506                  */
1507                 if ((one->mode ^ two->mode) & S_IFMT)
1508                         goto free_ab_and_return;
1509                 if (complete_rewrite &&
1510                     (textconv_one || !diff_filespec_is_binary(one)) &&
1511                     (textconv_two || !diff_filespec_is_binary(two))) {
1512                         emit_rewrite_diff(name_a, name_b, one, two,
1513                                                 textconv_one, textconv_two, o);
1514                         o->found_changes = 1;
1515                         goto free_ab_and_return;
1516                 }
1517         }
1519         if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1520                 die("unable to read files to diff");
1522         if (!DIFF_OPT_TST(o, TEXT) &&
1523             ( (diff_filespec_is_binary(one) && !textconv_one) ||
1524               (diff_filespec_is_binary(two) && !textconv_two) )) {
1525                 /* Quite common confusing case */
1526                 if (mf1.size == mf2.size &&
1527                     !memcmp(mf1.ptr, mf2.ptr, mf1.size))
1528                         goto free_ab_and_return;
1529                 if (DIFF_OPT_TST(o, BINARY))
1530                         emit_binary_diff(o->file, &mf1, &mf2);
1531                 else
1532                         fprintf(o->file, "Binary files %s and %s differ\n",
1533                                 lbl[0], lbl[1]);
1534                 o->found_changes = 1;
1535         }
1536         else {
1537                 /* Crazy xdl interfaces.. */
1538                 const char *diffopts = getenv("GIT_DIFF_OPTS");
1539                 xpparam_t xpp;
1540                 xdemitconf_t xecfg;
1541                 xdemitcb_t ecb;
1542                 struct emit_callback ecbdata;
1543                 const struct userdiff_funcname *pe;
1545                 if (textconv_one) {
1546                         size_t size;
1547                         mf1.ptr = run_textconv(textconv_one, one, &size);
1548                         if (!mf1.ptr)
1549                                 die("unable to read files to diff");
1550                         mf1.size = size;
1551                 }
1552                 if (textconv_two) {
1553                         size_t size;
1554                         mf2.ptr = run_textconv(textconv_two, two, &size);
1555                         if (!mf2.ptr)
1556                                 die("unable to read files to diff");
1557                         mf2.size = size;
1558                 }
1560                 pe = diff_funcname_pattern(one);
1561                 if (!pe)
1562                         pe = diff_funcname_pattern(two);
1564                 memset(&xpp, 0, sizeof(xpp));
1565                 memset(&xecfg, 0, sizeof(xecfg));
1566                 memset(&ecbdata, 0, sizeof(ecbdata));
1567                 ecbdata.label_path = lbl;
1568                 ecbdata.color_diff = DIFF_OPT_TST(o, COLOR_DIFF);
1569                 ecbdata.found_changesp = &o->found_changes;
1570                 ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
1571                 ecbdata.file = o->file;
1572                 xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1573                 xecfg.ctxlen = o->context;
1574                 xecfg.interhunkctxlen = o->interhunkcontext;
1575                 xecfg.flags = XDL_EMIT_FUNCNAMES;
1576                 if (pe)
1577                         xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
1578                 if (!diffopts)
1579                         ;
1580                 else if (!prefixcmp(diffopts, "--unified="))
1581                         xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
1582                 else if (!prefixcmp(diffopts, "-u"))
1583                         xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
1584                 if (DIFF_OPT_TST(o, COLOR_DIFF_WORDS)) {
1585                         ecbdata.diff_words =
1586                                 xcalloc(1, sizeof(struct diff_words_data));
1587                         ecbdata.diff_words->file = o->file;
1588                         if (!o->word_regex)
1589                                 o->word_regex = userdiff_word_regex(one);
1590                         if (!o->word_regex)
1591                                 o->word_regex = userdiff_word_regex(two);
1592                         if (!o->word_regex)
1593                                 o->word_regex = diff_word_regex_cfg;
1594                         if (o->word_regex) {
1595                                 ecbdata.diff_words->word_regex = (regex_t *)
1596                                         xmalloc(sizeof(regex_t));
1597                                 if (regcomp(ecbdata.diff_words->word_regex,
1598                                                 o->word_regex,
1599                                                 REG_EXTENDED | REG_NEWLINE))
1600                                         die ("Invalid regular expression: %s",
1601                                                         o->word_regex);
1602                         }
1603                 }
1604                 xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
1605                               &xpp, &xecfg, &ecb);
1606                 if (DIFF_OPT_TST(o, COLOR_DIFF_WORDS))
1607                         free_diff_words_data(&ecbdata);
1608                 if (textconv_one)
1609                         free(mf1.ptr);
1610                 if (textconv_two)
1611                         free(mf2.ptr);
1612                 xdiff_clear_find_func(&xecfg);
1613         }
1615  free_ab_and_return:
1616         diff_free_filespec_data(one);
1617         diff_free_filespec_data(two);
1618         free(a_one);
1619         free(b_two);
1620         return;
1623 static void builtin_diffstat(const char *name_a, const char *name_b,
1624                              struct diff_filespec *one,
1625                              struct diff_filespec *two,
1626                              struct diffstat_t *diffstat,
1627                              struct diff_options *o,
1628                              int complete_rewrite)
1630         mmfile_t mf1, mf2;
1631         struct diffstat_file *data;
1633         data = diffstat_add(diffstat, name_a, name_b);
1635         if (!one || !two) {
1636                 data->is_unmerged = 1;
1637                 return;
1638         }
1639         if (complete_rewrite) {
1640                 diff_populate_filespec(one, 0);
1641                 diff_populate_filespec(two, 0);
1642                 data->deleted = count_lines(one->data, one->size);
1643                 data->added = count_lines(two->data, two->size);
1644                 goto free_and_return;
1645         }
1646         if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1647                 die("unable to read files to diff");
1649         if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
1650                 data->is_binary = 1;
1651                 data->added = mf2.size;
1652                 data->deleted = mf1.size;
1653         } else {
1654                 /* Crazy xdl interfaces.. */
1655                 xpparam_t xpp;
1656                 xdemitconf_t xecfg;
1657                 xdemitcb_t ecb;
1659                 memset(&xpp, 0, sizeof(xpp));
1660                 memset(&xecfg, 0, sizeof(xecfg));
1661                 xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
1662                 xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
1663                               &xpp, &xecfg, &ecb);
1664         }
1666  free_and_return:
1667         diff_free_filespec_data(one);
1668         diff_free_filespec_data(two);
1671 static void builtin_checkdiff(const char *name_a, const char *name_b,
1672                               const char *attr_path,
1673                               struct diff_filespec *one,
1674                               struct diff_filespec *two,
1675                               struct diff_options *o)
1677         mmfile_t mf1, mf2;
1678         struct checkdiff_t data;
1680         if (!two)
1681                 return;
1683         memset(&data, 0, sizeof(data));
1684         data.filename = name_b ? name_b : name_a;
1685         data.lineno = 0;
1686         data.o = o;
1687         data.ws_rule = whitespace_rule(attr_path);
1689         if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
1690                 die("unable to read files to diff");
1692         /*
1693          * All the other codepaths check both sides, but not checking
1694          * the "old" side here is deliberate.  We are checking the newly
1695          * introduced changes, and as long as the "new" side is text, we
1696          * can and should check what it introduces.
1697          */
1698         if (diff_filespec_is_binary(two))
1699                 goto free_and_return;
1700         else {
1701                 /* Crazy xdl interfaces.. */
1702                 xpparam_t xpp;
1703                 xdemitconf_t xecfg;
1704                 xdemitcb_t ecb;
1706                 memset(&xpp, 0, sizeof(xpp));
1707                 memset(&xecfg, 0, sizeof(xecfg));
1708                 xecfg.ctxlen = 1; /* at least one context line */
1709                 xpp.flags = XDF_NEED_MINIMAL;
1710                 xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
1711                               &xpp, &xecfg, &ecb);
1713                 if ((data.ws_rule & WS_TRAILING_SPACE) &&
1714                     data.trailing_blanks_start) {
1715                         fprintf(o->file, "%s:%d: ends with blank lines.\n",
1716                                 data.filename, data.trailing_blanks_start);
1717                         data.status = 1; /* report errors */
1718                 }
1719         }
1720  free_and_return:
1721         diff_free_filespec_data(one);
1722         diff_free_filespec_data(two);
1723         if (data.status)
1724                 DIFF_OPT_SET(o, CHECK_FAILED);
1727 struct diff_filespec *alloc_filespec(const char *path)
1729         int namelen = strlen(path);
1730         struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
1732         memset(spec, 0, sizeof(*spec));
1733         spec->path = (char *)(spec + 1);
1734         memcpy(spec->path, path, namelen+1);
1735         spec->count = 1;
1736         spec->is_binary = -1;
1737         return spec;
1740 void free_filespec(struct diff_filespec *spec)
1742         if (!--spec->count) {
1743                 diff_free_filespec_data(spec);
1744                 free(spec);
1745         }
1748 void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
1749                    unsigned short mode)
1751         if (mode) {
1752                 spec->mode = canon_mode(mode);
1753                 hashcpy(spec->sha1, sha1);
1754                 spec->sha1_valid = !is_null_sha1(sha1);
1755         }
1758 /*
1759  * Given a name and sha1 pair, if the index tells us the file in
1760  * the work tree has that object contents, return true, so that
1761  * prepare_temp_file() does not have to inflate and extract.
1762  */
1763 static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
1765         struct cache_entry *ce;
1766         struct stat st;
1767         int pos, len;
1769         /*
1770          * We do not read the cache ourselves here, because the
1771          * benchmark with my previous version that always reads cache
1772          * shows that it makes things worse for diff-tree comparing
1773          * two linux-2.6 kernel trees in an already checked out work
1774          * tree.  This is because most diff-tree comparisons deal with
1775          * only a small number of files, while reading the cache is
1776          * expensive for a large project, and its cost outweighs the
1777          * savings we get by not inflating the object to a temporary
1778          * file.  Practically, this code only helps when we are used
1779          * by diff-cache --cached, which does read the cache before
1780          * calling us.
1781          */
1782         if (!active_cache)
1783                 return 0;
1785         /* We want to avoid the working directory if our caller
1786          * doesn't need the data in a normal file, this system
1787          * is rather slow with its stat/open/mmap/close syscalls,
1788          * and the object is contained in a pack file.  The pack
1789          * is probably already open and will be faster to obtain
1790          * the data through than the working directory.  Loose
1791          * objects however would tend to be slower as they need
1792          * to be individually opened and inflated.
1793          */
1794         if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
1795                 return 0;
1797         len = strlen(name);
1798         pos = cache_name_pos(name, len);
1799         if (pos < 0)
1800                 return 0;
1801         ce = active_cache[pos];
1803         /*
1804          * This is not the sha1 we are looking for, or
1805          * unreusable because it is not a regular file.
1806          */
1807         if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
1808                 return 0;
1810         /*
1811          * If ce is marked as "assume unchanged", there is no
1812          * guarantee that work tree matches what we are looking for.
1813          */
1814         if (ce->ce_flags & CE_VALID)
1815                 return 0;
1817         /*
1818          * If ce matches the file in the work tree, we can reuse it.
1819          */
1820         if (ce_uptodate(ce) ||
1821             (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
1822                 return 1;
1824         return 0;
1827 static int populate_from_stdin(struct diff_filespec *s)
1829         struct strbuf buf = STRBUF_INIT;
1830         size_t size = 0;
1832         if (strbuf_read(&buf, 0, 0) < 0)
1833                 return error("error while reading from stdin %s",
1834                                      strerror(errno));
1836         s->should_munmap = 0;
1837         s->data = strbuf_detach(&buf, &size);
1838         s->size = size;
1839         s->should_free = 1;
1840         return 0;
1843 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
1845         int len;
1846         char *data = xmalloc(100);
1847         len = snprintf(data, 100,
1848                 "Subproject commit %s\n", sha1_to_hex(s->sha1));
1849         s->data = data;
1850         s->size = len;
1851         s->should_free = 1;
1852         if (size_only) {
1853                 s->data = NULL;
1854                 free(data);
1855         }
1856         return 0;
1859 /*
1860  * While doing rename detection and pickaxe operation, we may need to
1861  * grab the data for the blob (or file) for our own in-core comparison.
1862  * diff_filespec has data and size fields for this purpose.
1863  */
1864 int diff_populate_filespec(struct diff_filespec *s, int size_only)
1866         int err = 0;
1867         if (!DIFF_FILE_VALID(s))
1868                 die("internal error: asking to populate invalid file.");
1869         if (S_ISDIR(s->mode))
1870                 return -1;
1872         if (s->data)
1873                 return 0;
1875         if (size_only && 0 < s->size)
1876                 return 0;
1878         if (S_ISGITLINK(s->mode))
1879                 return diff_populate_gitlink(s, size_only);
1881         if (!s->sha1_valid ||
1882             reuse_worktree_file(s->path, s->sha1, 0)) {
1883                 struct strbuf buf = STRBUF_INIT;
1884                 struct stat st;
1885                 int fd;
1887                 if (!strcmp(s->path, "-"))
1888                         return populate_from_stdin(s);
1890                 if (lstat(s->path, &st) < 0) {
1891                         if (errno == ENOENT) {
1892                         err_empty:
1893                                 err = -1;
1894                         empty:
1895                                 s->data = (char *)"";
1896                                 s->size = 0;
1897                                 return err;
1898                         }
1899                 }
1900                 s->size = xsize_t(st.st_size);
1901                 if (!s->size)
1902                         goto empty;
1903                 if (S_ISLNK(st.st_mode)) {
1904                         struct strbuf sb = STRBUF_INIT;
1906                         if (strbuf_readlink(&sb, s->path, s->size))
1907                                 goto err_empty;
1908                         s->size = sb.len;
1909                         s->data = strbuf_detach(&sb, NULL);
1910                         s->should_free = 1;
1911                         return 0;
1912                 }
1913                 if (size_only)
1914                         return 0;
1915                 fd = open(s->path, O_RDONLY);
1916                 if (fd < 0)
1917                         goto err_empty;
1918                 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
1919                 close(fd);
1920                 s->should_munmap = 1;
1922                 /*
1923                  * Convert from working tree format to canonical git format
1924                  */
1925                 if (convert_to_git(s->path, s->data, s->size, &buf, safe_crlf)) {
1926                         size_t size = 0;
1927                         munmap(s->data, s->size);
1928                         s->should_munmap = 0;
1929                         s->data = strbuf_detach(&buf, &size);
1930                         s->size = size;
1931                         s->should_free = 1;
1932                 }
1933         }
1934         else {
1935                 enum object_type type;
1936                 if (size_only)
1937                         type = sha1_object_info(s->sha1, &s->size);
1938                 else {
1939                         s->data = read_sha1_file(s->sha1, &type, &s->size);
1940                         s->should_free = 1;
1941                 }
1942         }
1943         return 0;
1946 void diff_free_filespec_blob(struct diff_filespec *s)
1948         if (s->should_free)
1949                 free(s->data);
1950         else if (s->should_munmap)
1951                 munmap(s->data, s->size);
1953         if (s->should_free || s->should_munmap) {
1954                 s->should_free = s->should_munmap = 0;
1955                 s->data = NULL;
1956         }
1959 void diff_free_filespec_data(struct diff_filespec *s)
1961         diff_free_filespec_blob(s);
1962         free(s->cnt_data);
1963         s->cnt_data = NULL;
1966 static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
1967                            void *blob,
1968                            unsigned long size,
1969                            const unsigned char *sha1,
1970                            int mode)
1972         int fd;
1973         struct strbuf buf = STRBUF_INIT;
1974         struct strbuf template = STRBUF_INIT;
1975         char *path_dup = xstrdup(path);
1976         const char *base = basename(path_dup);
1978         /* Generate "XXXXXX_basename.ext" */
1979         strbuf_addstr(&template, "XXXXXX_");
1980         strbuf_addstr(&template, base);
1982         fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
1983                         strlen(base) + 1);
1984         if (fd < 0)
1985                 die_errno("unable to create temp-file");
1986         if (convert_to_working_tree(path,
1987                         (const char *)blob, (size_t)size, &buf)) {
1988                 blob = buf.buf;
1989                 size = buf.len;
1990         }
1991         if (write_in_full(fd, blob, size) != size)
1992                 die_errno("unable to write temp-file");
1993         close(fd);
1994         temp->name = temp->tmp_path;
1995         strcpy(temp->hex, sha1_to_hex(sha1));
1996         temp->hex[40] = 0;
1997         sprintf(temp->mode, "%06o", mode);
1998         strbuf_release(&buf);
1999         strbuf_release(&template);
2000         free(path_dup);
2003 static struct diff_tempfile *prepare_temp_file(const char *name,
2004                 struct diff_filespec *one)
2006         struct diff_tempfile *temp = claim_diff_tempfile();
2008         if (!DIFF_FILE_VALID(one)) {
2009         not_a_valid_file:
2010                 /* A '-' entry produces this for file-2, and
2011                  * a '+' entry produces this for file-1.
2012                  */
2013                 temp->name = "/dev/null";
2014                 strcpy(temp->hex, ".");
2015                 strcpy(temp->mode, ".");
2016                 return temp;
2017         }
2019         if (!remove_tempfile_installed) {
2020                 atexit(remove_tempfile);
2021                 sigchain_push_common(remove_tempfile_on_signal);
2022                 remove_tempfile_installed = 1;
2023         }
2025         if (!one->sha1_valid ||
2026             reuse_worktree_file(name, one->sha1, 1)) {
2027                 struct stat st;
2028                 if (lstat(name, &st) < 0) {
2029                         if (errno == ENOENT)
2030                                 goto not_a_valid_file;
2031                         die_errno("stat(%s)", name);
2032                 }
2033                 if (S_ISLNK(st.st_mode)) {
2034                         struct strbuf sb = STRBUF_INIT;
2035                         if (strbuf_readlink(&sb, name, st.st_size) < 0)
2036                                 die_errno("readlink(%s)", name);
2037                         prep_temp_blob(name, temp, sb.buf, sb.len,
2038                                        (one->sha1_valid ?
2039                                         one->sha1 : null_sha1),
2040                                        (one->sha1_valid ?
2041                                         one->mode : S_IFLNK));
2042                         strbuf_release(&sb);
2043                 }
2044                 else {
2045                         /* we can borrow from the file in the work tree */
2046                         temp->name = name;
2047                         if (!one->sha1_valid)
2048                                 strcpy(temp->hex, sha1_to_hex(null_sha1));
2049                         else
2050                                 strcpy(temp->hex, sha1_to_hex(one->sha1));
2051                         /* Even though we may sometimes borrow the
2052                          * contents from the work tree, we always want
2053                          * one->mode.  mode is trustworthy even when
2054                          * !(one->sha1_valid), as long as
2055                          * DIFF_FILE_VALID(one).
2056                          */
2057                         sprintf(temp->mode, "%06o", one->mode);
2058                 }
2059                 return temp;
2060         }
2061         else {
2062                 if (diff_populate_filespec(one, 0))
2063                         die("cannot read data blob for %s", one->path);
2064                 prep_temp_blob(name, temp, one->data, one->size,
2065                                one->sha1, one->mode);
2066         }
2067         return temp;
2070 /* An external diff command takes:
2071  *
2072  * diff-cmd name infile1 infile1-sha1 infile1-mode \
2073  *               infile2 infile2-sha1 infile2-mode [ rename-to ]
2074  *
2075  */
2076 static void run_external_diff(const char *pgm,
2077                               const char *name,
2078                               const char *other,
2079                               struct diff_filespec *one,
2080                               struct diff_filespec *two,
2081                               const char *xfrm_msg,
2082                               int complete_rewrite)
2084         const char *spawn_arg[10];
2085         int retval;
2086         const char **arg = &spawn_arg[0];
2088         if (one && two) {
2089                 struct diff_tempfile *temp_one, *temp_two;
2090                 const char *othername = (other ? other : name);
2091                 temp_one = prepare_temp_file(name, one);
2092                 temp_two = prepare_temp_file(othername, two);
2093                 *arg++ = pgm;
2094                 *arg++ = name;
2095                 *arg++ = temp_one->name;
2096                 *arg++ = temp_one->hex;
2097                 *arg++ = temp_one->mode;
2098                 *arg++ = temp_two->name;
2099                 *arg++ = temp_two->hex;
2100                 *arg++ = temp_two->mode;
2101                 if (other) {
2102                         *arg++ = other;
2103                         *arg++ = xfrm_msg;
2104                 }
2105         } else {
2106                 *arg++ = pgm;
2107                 *arg++ = name;
2108         }
2109         *arg = NULL;
2110         fflush(NULL);
2111         retval = run_command_v_opt(spawn_arg, 0);
2112         remove_tempfile();
2113         if (retval) {
2114                 fprintf(stderr, "external diff died, stopping at %s.\n", name);
2115                 exit(1);
2116         }
2119 static int similarity_index(struct diff_filepair *p)
2121         return p->score * 100 / MAX_SCORE;
2124 static void fill_metainfo(struct strbuf *msg,
2125                           const char *name,
2126                           const char *other,
2127                           struct diff_filespec *one,
2128                           struct diff_filespec *two,
2129                           struct diff_options *o,
2130                           struct diff_filepair *p)
2132         strbuf_init(msg, PATH_MAX * 2 + 300);
2133         switch (p->status) {
2134         case DIFF_STATUS_COPIED:
2135                 strbuf_addf(msg, "similarity index %d%%", similarity_index(p));
2136                 strbuf_addstr(msg, "\ncopy from ");
2137                 quote_c_style(name, msg, NULL, 0);
2138                 strbuf_addstr(msg, "\ncopy to ");
2139                 quote_c_style(other, msg, NULL, 0);
2140                 strbuf_addch(msg, '\n');
2141                 break;
2142         case DIFF_STATUS_RENAMED:
2143                 strbuf_addf(msg, "similarity index %d%%", similarity_index(p));
2144                 strbuf_addstr(msg, "\nrename from ");
2145                 quote_c_style(name, msg, NULL, 0);
2146                 strbuf_addstr(msg, "\nrename to ");
2147                 quote_c_style(other, msg, NULL, 0);
2148                 strbuf_addch(msg, '\n');
2149                 break;
2150         case DIFF_STATUS_MODIFIED:
2151                 if (p->score) {
2152                         strbuf_addf(msg, "dissimilarity index %d%%\n",
2153                                     similarity_index(p));
2154                         break;
2155                 }
2156                 /* fallthru */
2157         default:
2158                 /* nothing */
2159                 ;
2160         }
2161         if (one && two && hashcmp(one->sha1, two->sha1)) {
2162                 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2164                 if (DIFF_OPT_TST(o, BINARY)) {
2165                         mmfile_t mf;
2166                         if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
2167                             (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
2168                                 abbrev = 40;
2169                 }
2170                 strbuf_addf(msg, "index %.*s..%.*s",
2171                             abbrev, sha1_to_hex(one->sha1),
2172                             abbrev, sha1_to_hex(two->sha1));
2173                 if (one->mode == two->mode)
2174                         strbuf_addf(msg, " %06o", one->mode);
2175                 strbuf_addch(msg, '\n');
2176         }
2177         if (msg->len)
2178                 strbuf_setlen(msg, msg->len - 1);
2181 static void run_diff_cmd(const char *pgm,
2182                          const char *name,
2183                          const char *other,
2184                          const char *attr_path,
2185                          struct diff_filespec *one,
2186                          struct diff_filespec *two,
2187                          struct strbuf *msg,
2188                          struct diff_options *o,
2189                          struct diff_filepair *p)
2191         const char *xfrm_msg = NULL;
2192         int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
2194         if (msg) {
2195                 fill_metainfo(msg, name, other, one, two, o, p);
2196                 xfrm_msg = msg->len ? msg->buf : NULL;
2197         }
2199         if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
2200                 pgm = NULL;
2201         else {
2202                 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
2203                 if (drv && drv->external)
2204                         pgm = drv->external;
2205         }
2207         if (pgm) {
2208                 run_external_diff(pgm, name, other, one, two, xfrm_msg,
2209                                   complete_rewrite);
2210                 return;
2211         }
2212         if (one && two)
2213                 builtin_diff(name, other ? other : name,
2214                              one, two, xfrm_msg, o, complete_rewrite);
2215         else
2216                 fprintf(o->file, "* Unmerged path %s\n", name);
2219 static void diff_fill_sha1_info(struct diff_filespec *one)
2221         if (DIFF_FILE_VALID(one)) {
2222                 if (!one->sha1_valid) {
2223                         struct stat st;
2224                         if (!strcmp(one->path, "-")) {
2225                                 hashcpy(one->sha1, null_sha1);
2226                                 return;
2227                         }
2228                         if (lstat(one->path, &st) < 0)
2229                                 die_errno("stat '%s'", one->path);
2230                         if (index_path(one->sha1, one->path, &st, 0))
2231                                 die("cannot hash %s", one->path);
2232                 }
2233         }
2234         else
2235                 hashclr(one->sha1);
2238 static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
2240         /* Strip the prefix but do not molest /dev/null and absolute paths */
2241         if (*namep && **namep != '/')
2242                 *namep += prefix_length;
2243         if (*otherp && **otherp != '/')
2244                 *otherp += prefix_length;
2247 static void run_diff(struct diff_filepair *p, struct diff_options *o)
2249         const char *pgm = external_diff();
2250         struct strbuf msg;
2251         struct diff_filespec *one = p->one;
2252         struct diff_filespec *two = p->two;
2253         const char *name;
2254         const char *other;
2255         const char *attr_path;
2257         name  = p->one->path;
2258         other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2259         attr_path = name;
2260         if (o->prefix_length)
2261                 strip_prefix(o->prefix_length, &name, &other);
2263         if (DIFF_PAIR_UNMERGED(p)) {
2264                 run_diff_cmd(pgm, name, NULL, attr_path,
2265                              NULL, NULL, NULL, o, p);
2266                 return;
2267         }
2269         diff_fill_sha1_info(one);
2270         diff_fill_sha1_info(two);
2272         if (!pgm &&
2273             DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
2274             (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
2275                 /*
2276                  * a filepair that changes between file and symlink
2277                  * needs to be split into deletion and creation.
2278                  */
2279                 struct diff_filespec *null = alloc_filespec(two->path);
2280                 run_diff_cmd(NULL, name, other, attr_path,
2281                              one, null, &msg, o, p);
2282                 free(null);
2283                 strbuf_release(&msg);
2285                 null = alloc_filespec(one->path);
2286                 run_diff_cmd(NULL, name, other, attr_path,
2287                              null, two, &msg, o, p);
2288                 free(null);
2289         }
2290         else
2291                 run_diff_cmd(pgm, name, other, attr_path,
2292                              one, two, &msg, o, p);
2294         strbuf_release(&msg);
2297 static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
2298                          struct diffstat_t *diffstat)
2300         const char *name;
2301         const char *other;
2302         int complete_rewrite = 0;
2304         if (DIFF_PAIR_UNMERGED(p)) {
2305                 /* unmerged */
2306                 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
2307                 return;
2308         }
2310         name = p->one->path;
2311         other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2313         if (o->prefix_length)
2314                 strip_prefix(o->prefix_length, &name, &other);
2316         diff_fill_sha1_info(p->one);
2317         diff_fill_sha1_info(p->two);
2319         if (p->status == DIFF_STATUS_MODIFIED && p->score)
2320                 complete_rewrite = 1;
2321         builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
2324 static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
2326         const char *name;
2327         const char *other;
2328         const char *attr_path;
2330         if (DIFF_PAIR_UNMERGED(p)) {
2331                 /* unmerged */
2332                 return;
2333         }
2335         name = p->one->path;
2336         other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2337         attr_path = other ? other : name;
2339         if (o->prefix_length)
2340                 strip_prefix(o->prefix_length, &name, &other);
2342         diff_fill_sha1_info(p->one);
2343         diff_fill_sha1_info(p->two);
2345         builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
2348 void diff_setup(struct diff_options *options)
2350         memset(options, 0, sizeof(*options));
2352         options->file = stdout;
2354         options->line_termination = '\n';
2355         options->break_opt = -1;
2356         options->rename_limit = -1;
2357         options->dirstat_percent = 3;
2358         options->context = 3;
2360         options->change = diff_change;
2361         options->add_remove = diff_addremove;
2362         if (diff_use_color_default > 0)
2363                 DIFF_OPT_SET(options, COLOR_DIFF);
2364         options->detect_rename = diff_detect_rename_default;
2366         if (!diff_mnemonic_prefix) {
2367                 options->a_prefix = "a/";
2368                 options->b_prefix = "b/";
2369         }
2372 int diff_setup_done(struct diff_options *options)
2374         int count = 0;
2376         if (options->output_format & DIFF_FORMAT_NAME)
2377                 count++;
2378         if (options->output_format & DIFF_FORMAT_NAME_STATUS)
2379                 count++;
2380         if (options->output_format & DIFF_FORMAT_CHECKDIFF)
2381                 count++;
2382         if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
2383                 count++;
2384         if (count > 1)
2385                 die("--name-only, --name-status, --check and -s are mutually exclusive");
2387         if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
2388                 options->detect_rename = DIFF_DETECT_COPY;
2390         if (!DIFF_OPT_TST(options, RELATIVE_NAME))
2391                 options->prefix = NULL;
2392         if (options->prefix)
2393                 options->prefix_length = strlen(options->prefix);
2394         else
2395                 options->prefix_length = 0;
2397         if (options->output_format & (DIFF_FORMAT_NAME |
2398                                       DIFF_FORMAT_NAME_STATUS |
2399                                       DIFF_FORMAT_CHECKDIFF |
2400                                       DIFF_FORMAT_NO_OUTPUT))
2401                 options->output_format &= ~(DIFF_FORMAT_RAW |
2402                                             DIFF_FORMAT_NUMSTAT |
2403                                             DIFF_FORMAT_DIFFSTAT |
2404                                             DIFF_FORMAT_SHORTSTAT |
2405                                             DIFF_FORMAT_DIRSTAT |
2406                                             DIFF_FORMAT_SUMMARY |
2407                                             DIFF_FORMAT_PATCH);
2409         /*
2410          * These cases always need recursive; we do not drop caller-supplied
2411          * recursive bits for other formats here.
2412          */
2413         if (options->output_format & (DIFF_FORMAT_PATCH |
2414                                       DIFF_FORMAT_NUMSTAT |
2415                                       DIFF_FORMAT_DIFFSTAT |
2416                                       DIFF_FORMAT_SHORTSTAT |
2417                                       DIFF_FORMAT_DIRSTAT |
2418                                       DIFF_FORMAT_SUMMARY |
2419                                       DIFF_FORMAT_CHECKDIFF))
2420                 DIFF_OPT_SET(options, RECURSIVE);
2421         /*
2422          * Also pickaxe would not work very well if you do not say recursive
2423          */
2424         if (options->pickaxe)
2425                 DIFF_OPT_SET(options, RECURSIVE);
2427         if (options->detect_rename && options->rename_limit < 0)
2428                 options->rename_limit = diff_rename_limit_default;
2429         if (options->setup & DIFF_SETUP_USE_CACHE) {
2430                 if (!active_cache)
2431                         /* read-cache does not die even when it fails
2432                          * so it is safe for us to do this here.  Also
2433                          * it does not smudge active_cache or active_nr
2434                          * when it fails, so we do not have to worry about
2435                          * cleaning it up ourselves either.
2436                          */
2437                         read_cache();
2438         }
2439         if (options->abbrev <= 0 || 40 < options->abbrev)
2440                 options->abbrev = 40; /* full */
2442         /*
2443          * It does not make sense to show the first hit we happened
2444          * to have found.  It does not make sense not to return with
2445          * exit code in such a case either.
2446          */
2447         if (DIFF_OPT_TST(options, QUIET)) {
2448                 options->output_format = DIFF_FORMAT_NO_OUTPUT;
2449                 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
2450         }
2452         return 0;
2455 static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
2457         char c, *eq;
2458         int len;
2460         if (*arg != '-')
2461                 return 0;
2462         c = *++arg;
2463         if (!c)
2464                 return 0;
2465         if (c == arg_short) {
2466                 c = *++arg;
2467                 if (!c)
2468                         return 1;
2469                 if (val && isdigit(c)) {
2470                         char *end;
2471                         int n = strtoul(arg, &end, 10);
2472                         if (*end)
2473                                 return 0;
2474                         *val = n;
2475                         return 1;
2476                 }
2477                 return 0;
2478         }
2479         if (c != '-')
2480                 return 0;
2481         arg++;
2482         eq = strchr(arg, '=');
2483         if (eq)
2484                 len = eq - arg;
2485         else
2486                 len = strlen(arg);
2487         if (!len || strncmp(arg, arg_long, len))
2488                 return 0;
2489         if (eq) {
2490                 int n;
2491                 char *end;
2492                 if (!isdigit(*++eq))
2493                         return 0;
2494                 n = strtoul(eq, &end, 10);
2495                 if (*end)
2496                         return 0;
2497                 *val = n;
2498         }
2499         return 1;
2502 static int diff_scoreopt_parse(const char *opt);
2504 int diff_opt_parse(struct diff_options *options, const char **av, int ac)
2506         const char *arg = av[0];
2508         /* Output format options */
2509         if (!strcmp(arg, "-p") || !strcmp(arg, "-u"))
2510                 options->output_format |= DIFF_FORMAT_PATCH;
2511         else if (opt_arg(arg, 'U', "unified", &options->context))
2512                 options->output_format |= DIFF_FORMAT_PATCH;
2513         else if (!strcmp(arg, "--raw"))
2514                 options->output_format |= DIFF_FORMAT_RAW;
2515         else if (!strcmp(arg, "--patch-with-raw"))
2516                 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
2517         else if (!strcmp(arg, "--numstat"))
2518                 options->output_format |= DIFF_FORMAT_NUMSTAT;
2519         else if (!strcmp(arg, "--shortstat"))
2520                 options->output_format |= DIFF_FORMAT_SHORTSTAT;
2521         else if (opt_arg(arg, 'X', "dirstat", &options->dirstat_percent))
2522                 options->output_format |= DIFF_FORMAT_DIRSTAT;
2523         else if (!strcmp(arg, "--cumulative")) {
2524                 options->output_format |= DIFF_FORMAT_DIRSTAT;
2525                 DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
2526         } else if (opt_arg(arg, 0, "dirstat-by-file",
2527                            &options->dirstat_percent)) {
2528                 options->output_format |= DIFF_FORMAT_DIRSTAT;
2529                 DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
2530         }
2531         else if (!strcmp(arg, "--check"))
2532                 options->output_format |= DIFF_FORMAT_CHECKDIFF;
2533         else if (!strcmp(arg, "--summary"))
2534                 options->output_format |= DIFF_FORMAT_SUMMARY;
2535         else if (!strcmp(arg, "--patch-with-stat"))
2536                 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
2537         else if (!strcmp(arg, "--name-only"))
2538                 options->output_format |= DIFF_FORMAT_NAME;
2539         else if (!strcmp(arg, "--name-status"))
2540                 options->output_format |= DIFF_FORMAT_NAME_STATUS;
2541         else if (!strcmp(arg, "-s"))
2542                 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
2543         else if (!prefixcmp(arg, "--stat")) {
2544                 char *end;
2545                 int width = options->stat_width;
2546                 int name_width = options->stat_name_width;
2547                 arg += 6;
2548                 end = (char *)arg;
2550                 switch (*arg) {
2551                 case '-':
2552                         if (!prefixcmp(arg, "-width="))
2553                                 width = strtoul(arg + 7, &end, 10);
2554                         else if (!prefixcmp(arg, "-name-width="))
2555                                 name_width = strtoul(arg + 12, &end, 10);
2556                         break;
2557                 case '=':
2558                         width = strtoul(arg+1, &end, 10);
2559                         if (*end == ',')
2560                                 name_width = strtoul(end+1, &end, 10);
2561                 }
2563                 /* Important! This checks all the error cases! */
2564                 if (*end)
2565                         return 0;
2566                 options->output_format |= DIFF_FORMAT_DIFFSTAT;
2567                 options->stat_name_width = name_width;
2568                 options->stat_width = width;
2569         }
2571         /* renames options */
2572         else if (!prefixcmp(arg, "-B")) {
2573                 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
2574                         return -1;
2575         }
2576         else if (!prefixcmp(arg, "-M")) {
2577                 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
2578                         return -1;
2579                 options->detect_rename = DIFF_DETECT_RENAME;
2580         }
2581         else if (!prefixcmp(arg, "-C")) {
2582                 if (options->detect_rename == DIFF_DETECT_COPY)
2583                         DIFF_OPT_SET(options, FIND_COPIES_HARDER);
2584                 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
2585                         return -1;
2586                 options->detect_rename = DIFF_DETECT_COPY;
2587         }
2588         else if (!strcmp(arg, "--no-renames"))
2589                 options->detect_rename = 0;
2590         else if (!strcmp(arg, "--relative"))
2591                 DIFF_OPT_SET(options, RELATIVE_NAME);
2592         else if (!prefixcmp(arg, "--relative=")) {
2593                 DIFF_OPT_SET(options, RELATIVE_NAME);
2594                 options->prefix = arg + 11;
2595         }
2597         /* xdiff options */
2598         else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
2599                 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
2600         else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
2601                 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
2602         else if (!strcmp(arg, "--ignore-space-at-eol"))
2603                 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
2604         else if (!strcmp(arg, "--patience"))
2605                 DIFF_XDL_SET(options, PATIENCE_DIFF);
2607         /* flags options */
2608         else if (!strcmp(arg, "--binary")) {
2609                 options->output_format |= DIFF_FORMAT_PATCH;
2610                 DIFF_OPT_SET(options, BINARY);
2611         }
2612         else if (!strcmp(arg, "--full-index"))
2613                 DIFF_OPT_SET(options, FULL_INDEX);
2614         else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
2615                 DIFF_OPT_SET(options, TEXT);
2616         else if (!strcmp(arg, "-R"))
2617                 DIFF_OPT_SET(options, REVERSE_DIFF);
2618         else if (!strcmp(arg, "--find-copies-harder"))
2619                 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
2620         else if (!strcmp(arg, "--follow"))
2621                 DIFF_OPT_SET(options, FOLLOW_RENAMES);
2622         else if (!strcmp(arg, "--color"))
2623                 DIFF_OPT_SET(options, COLOR_DIFF);
2624         else if (!strcmp(arg, "--no-color"))
2625                 DIFF_OPT_CLR(options, COLOR_DIFF);
2626         else if (!strcmp(arg, "--color-words")) {
2627                 DIFF_OPT_SET(options, COLOR_DIFF);
2628                 DIFF_OPT_SET(options, COLOR_DIFF_WORDS);
2629         }
2630         else if (!prefixcmp(arg, "--color-words=")) {
2631                 DIFF_OPT_SET(options, COLOR_DIFF);
2632                 DIFF_OPT_SET(options, COLOR_DIFF_WORDS);
2633                 options->word_regex = arg + 14;
2634         }
2635         else if (!strcmp(arg, "--exit-code"))
2636                 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
2637         else if (!strcmp(arg, "--quiet"))
2638                 DIFF_OPT_SET(options, QUIET);
2639         else if (!strcmp(arg, "--ext-diff"))
2640                 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
2641         else if (!strcmp(arg, "--no-ext-diff"))
2642                 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
2643         else if (!strcmp(arg, "--textconv"))
2644                 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
2645         else if (!strcmp(arg, "--no-textconv"))
2646                 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
2647         else if (!strcmp(arg, "--ignore-submodules"))
2648                 DIFF_OPT_SET(options, IGNORE_SUBMODULES);
2650         /* misc options */
2651         else if (!strcmp(arg, "-z"))
2652                 options->line_termination = 0;
2653         else if (!prefixcmp(arg, "-l"))
2654                 options->rename_limit = strtoul(arg+2, NULL, 10);
2655         else if (!prefixcmp(arg, "-S"))
2656                 options->pickaxe = arg + 2;
2657         else if (!strcmp(arg, "--pickaxe-all"))
2658                 options->pickaxe_opts = DIFF_PICKAXE_ALL;
2659         else if (!strcmp(arg, "--pickaxe-regex"))
2660                 options->pickaxe_opts = DIFF_PICKAXE_REGEX;
2661         else if (!prefixcmp(arg, "-O"))
2662                 options->orderfile = arg + 2;
2663         else if (!prefixcmp(arg, "--diff-filter="))
2664                 options->filter = arg + 14;
2665         else if (!strcmp(arg, "--abbrev"))
2666                 options->abbrev = DEFAULT_ABBREV;
2667         else if (!prefixcmp(arg, "--abbrev=")) {
2668                 options->abbrev = strtoul(arg + 9, NULL, 10);
2669                 if (options->abbrev < MINIMUM_ABBREV)
2670                         options->abbrev = MINIMUM_ABBREV;
2671                 else if (40 < options->abbrev)
2672                         options->abbrev = 40;
2673         }
2674         else if (!prefixcmp(arg, "--src-prefix="))
2675                 options->a_prefix = arg + 13;
2676         else if (!prefixcmp(arg, "--dst-prefix="))
2677                 options->b_prefix = arg + 13;
2678         else if (!strcmp(arg, "--no-prefix"))
2679                 options->a_prefix = options->b_prefix = "";
2680         else if (opt_arg(arg, '\0', "inter-hunk-context",
2681                          &options->interhunkcontext))
2682                 ;
2683         else if (!prefixcmp(arg, "--output=")) {
2684                 options->file = fopen(arg + strlen("--output="), "w");
2685                 options->close_file = 1;
2686         } else
2687                 return 0;
2688         return 1;
2691 static int parse_num(const char **cp_p)
2693         unsigned long num, scale;
2694         int ch, dot;
2695         const char *cp = *cp_p;
2697         num = 0;
2698         scale = 1;
2699         dot = 0;
2700         for (;;) {
2701                 ch = *cp;
2702                 if ( !dot && ch == '.' ) {
2703                         scale = 1;
2704                         dot = 1;
2705                 } else if ( ch == '%' ) {
2706                         scale = dot ? scale*100 : 100;
2707                         cp++;   /* % is always at the end */
2708                         break;
2709                 } else if ( ch >= '0' && ch <= '9' ) {
2710                         if ( scale < 100000 ) {
2711                                 scale *= 10;
2712                                 num = (num*10) + (ch-'0');
2713                         }
2714                 } else {
2715                         break;
2716                 }
2717                 cp++;
2718         }
2719         *cp_p = cp;
2721         /* user says num divided by scale and we say internally that
2722          * is MAX_SCORE * num / scale.
2723          */
2724         return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
2727 static int diff_scoreopt_parse(const char *opt)
2729         int opt1, opt2, cmd;
2731         if (*opt++ != '-')
2732                 return -1;
2733         cmd = *opt++;
2734         if (cmd != 'M' && cmd != 'C' && cmd != 'B')
2735                 return -1; /* that is not a -M, -C nor -B option */
2737         opt1 = parse_num(&opt);
2738         if (cmd != 'B')
2739                 opt2 = 0;
2740         else {
2741                 if (*opt == 0)
2742                         opt2 = 0;
2743                 else if (*opt != '/')
2744                         return -1; /* we expect -B80/99 or -B80 */
2745                 else {
2746                         opt++;
2747                         opt2 = parse_num(&opt);
2748                 }
2749         }
2750         if (*opt != 0)
2751                 return -1;
2752         return opt1 | (opt2 << 16);
2755 struct diff_queue_struct diff_queued_diff;
2757 void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
2759         if (queue->alloc <= queue->nr) {
2760                 queue->alloc = alloc_nr(queue->alloc);
2761                 queue->queue = xrealloc(queue->queue,
2762                                         sizeof(dp) * queue->alloc);
2763         }
2764         queue->queue[queue->nr++] = dp;
2767 struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
2768                                  struct diff_filespec *one,
2769                                  struct diff_filespec *two)
2771         struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
2772         dp->one = one;
2773         dp->two = two;
2774         if (queue)
2775                 diff_q(queue, dp);
2776         return dp;
2779 void diff_free_filepair(struct diff_filepair *p)
2781         free_filespec(p->one);
2782         free_filespec(p->two);
2783         free(p);
2786 /* This is different from find_unique_abbrev() in that
2787  * it stuffs the result with dots for alignment.
2788  */
2789 const char *diff_unique_abbrev(const unsigned char *sha1, int len)
2791         int abblen;
2792         const char *abbrev;
2793         if (len == 40)
2794                 return sha1_to_hex(sha1);
2796         abbrev = find_unique_abbrev(sha1, len);
2797         abblen = strlen(abbrev);
2798         if (abblen < 37) {
2799                 static char hex[41];
2800                 if (len < abblen && abblen <= len + 2)
2801                         sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
2802                 else
2803                         sprintf(hex, "%s...", abbrev);
2804                 return hex;
2805         }
2806         return sha1_to_hex(sha1);
2809 static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
2811         int line_termination = opt->line_termination;
2812         int inter_name_termination = line_termination ? '\t' : '\0';
2814         if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
2815                 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
2816                         diff_unique_abbrev(p->one->sha1, opt->abbrev));
2817                 fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
2818         }
2819         if (p->score) {
2820                 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
2821                         inter_name_termination);
2822         } else {
2823                 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
2824         }
2826         if (p->status == DIFF_STATUS_COPIED ||
2827             p->status == DIFF_STATUS_RENAMED) {
2828                 const char *name_a, *name_b;
2829                 name_a = p->one->path;
2830                 name_b = p->two->path;
2831                 strip_prefix(opt->prefix_length, &name_a, &name_b);
2832                 write_name_quoted(name_a, opt->file, inter_name_termination);
2833                 write_name_quoted(name_b, opt->file, line_termination);
2834         } else {
2835                 const char *name_a, *name_b;
2836                 name_a = p->one->mode ? p->one->path : p->two->path;
2837                 name_b = NULL;
2838                 strip_prefix(opt->prefix_length, &name_a, &name_b);
2839                 write_name_quoted(name_a, opt->file, line_termination);
2840         }
2843 int diff_unmodified_pair(struct diff_filepair *p)
2845         /* This function is written stricter than necessary to support
2846          * the currently implemented transformers, but the idea is to
2847          * let transformers to produce diff_filepairs any way they want,
2848          * and filter and clean them up here before producing the output.
2849          */
2850         struct diff_filespec *one = p->one, *two = p->two;
2852         if (DIFF_PAIR_UNMERGED(p))
2853                 return 0; /* unmerged is interesting */
2855         /* deletion, addition, mode or type change
2856          * and rename are all interesting.
2857          */
2858         if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
2859             DIFF_PAIR_MODE_CHANGED(p) ||
2860             strcmp(one->path, two->path))
2861                 return 0;
2863         /* both are valid and point at the same path.  that is, we are
2864          * dealing with a change.
2865          */
2866         if (one->sha1_valid && two->sha1_valid &&
2867             !hashcmp(one->sha1, two->sha1))
2868                 return 1; /* no change */
2869         if (!one->sha1_valid && !two->sha1_valid)
2870                 return 1; /* both look at the same file on the filesystem. */
2871         return 0;
2874 static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
2876         if (diff_unmodified_pair(p))
2877                 return;
2879         if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2880             (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2881                 return; /* no tree diffs in patch format */
2883         run_diff(p, o);
2886 static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
2887                             struct diffstat_t *diffstat)
2889         if (diff_unmodified_pair(p))
2890                 return;
2892         if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2893             (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2894                 return; /* no tree diffs in patch format */
2896         run_diffstat(p, o, diffstat);
2899 static void diff_flush_checkdiff(struct diff_filepair *p,
2900                 struct diff_options *o)
2902         if (diff_unmodified_pair(p))
2903                 return;
2905         if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
2906             (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
2907                 return; /* no tree diffs in patch format */
2909         run_checkdiff(p, o);
2912 int diff_queue_is_empty(void)
2914         struct diff_queue_struct *q = &diff_queued_diff;
2915         int i;
2916         for (i = 0; i < q->nr; i++)
2917                 if (!diff_unmodified_pair(q->queue[i]))
2918                         return 0;
2919         return 1;
2922 #if DIFF_DEBUG
2923 void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
2925         fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
2926                 x, one ? one : "",
2927                 s->path,
2928                 DIFF_FILE_VALID(s) ? "valid" : "invalid",
2929                 s->mode,
2930                 s->sha1_valid ? sha1_to_hex(s->sha1) : "");
2931         fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
2932                 x, one ? one : "",
2933                 s->size, s->xfrm_flags);
2936 void diff_debug_filepair(const struct diff_filepair *p, int i)
2938         diff_debug_filespec(p->one, i, "one");
2939         diff_debug_filespec(p->two, i, "two");
2940         fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
2941                 p->score, p->status ? p->status : '?',
2942                 p->one->rename_used, p->broken_pair);
2945 void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
2947         int i;
2948         if (msg)
2949                 fprintf(stderr, "%s\n", msg);
2950         fprintf(stderr, "q->nr = %d\n", q->nr);
2951         for (i = 0; i < q->nr; i++) {
2952                 struct diff_filepair *p = q->queue[i];
2953                 diff_debug_filepair(p, i);
2954         }
2956 #endif
2958 static void diff_resolve_rename_copy(void)
2960         int i;
2961         struct diff_filepair *p;
2962         struct diff_queue_struct *q = &diff_queued_diff;
2964         diff_debug_queue("resolve-rename-copy", q);
2966         for (i = 0; i < q->nr; i++) {
2967                 p = q->queue[i];
2968                 p->status = 0; /* undecided */
2969                 if (DIFF_PAIR_UNMERGED(p))
2970                         p->status = DIFF_STATUS_UNMERGED;
2971                 else if (!DIFF_FILE_VALID(p->one))
2972                         p->status = DIFF_STATUS_ADDED;
2973                 else if (!DIFF_FILE_VALID(p->two))
2974                         p->status = DIFF_STATUS_DELETED;
2975                 else if (DIFF_PAIR_TYPE_CHANGED(p))
2976                         p->status = DIFF_STATUS_TYPE_CHANGED;
2978                 /* from this point on, we are dealing with a pair
2979                  * whose both sides are valid and of the same type, i.e.
2980                  * either in-place edit or rename/copy edit.
2981                  */
2982                 else if (DIFF_PAIR_RENAME(p)) {
2983                         /*
2984                          * A rename might have re-connected a broken
2985                          * pair up, causing the pathnames to be the
2986                          * same again. If so, that's not a rename at
2987                          * all, just a modification..
2988                          *
2989                          * Otherwise, see if this source was used for
2990                          * multiple renames, in which case we decrement
2991                          * the count, and call it a copy.
2992                          */
2993                         if (!strcmp(p->one->path, p->two->path))
2994                                 p->status = DIFF_STATUS_MODIFIED;
2995                         else if (--p->one->rename_used > 0)
2996                                 p->status = DIFF_STATUS_COPIED;
2997                         else
2998                                 p->status = DIFF_STATUS_RENAMED;
2999                 }
3000                 else if (hashcmp(p->one->sha1, p->two->sha1) ||
3001                          p->one->mode != p->two->mode ||
3002                          is_null_sha1(p->one->sha1))
3003                         p->status = DIFF_STATUS_MODIFIED;
3004                 else {
3005                         /* This is a "no-change" entry and should not
3006                          * happen anymore, but prepare for broken callers.
3007                          */
3008                         error("feeding unmodified %s to diffcore",
3009                               p->one->path);
3010                         p->status = DIFF_STATUS_UNKNOWN;
3011                 }
3012         }
3013         diff_debug_queue("resolve-rename-copy done", q);
3016 static int check_pair_status(struct diff_filepair *p)
3018         switch (p->status) {
3019         case DIFF_STATUS_UNKNOWN:
3020                 return 0;
3021         case 0:
3022                 die("internal error in diff-resolve-rename-copy");
3023         default:
3024                 return 1;
3025         }
3028 static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
3030         int fmt = opt->output_format;
3032         if (fmt & DIFF_FORMAT_CHECKDIFF)
3033                 diff_flush_checkdiff(p, opt);
3034         else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
3035                 diff_flush_raw(p, opt);
3036         else if (fmt & DIFF_FORMAT_NAME) {
3037                 const char *name_a, *name_b;
3038                 name_a = p->two->path;
3039                 name_b = NULL;
3040                 strip_prefix(opt->prefix_length, &name_a, &name_b);
3041                 write_name_quoted(name_a, opt->file, opt->line_termination);
3042         }
3045 static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
3047         if (fs->mode)
3048                 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
3049         else
3050                 fprintf(file, " %s ", newdelete);
3051         write_name_quoted(fs->path, file, '\n');
3055 static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name)
3057         if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
3058                 fprintf(file, " mode change %06o => %06o%c", p->one->mode, p->two->mode,
3059                         show_name ? ' ' : '\n');
3060                 if (show_name) {
3061                         write_name_quoted(p->two->path, file, '\n');
3062                 }
3063         }
3066 static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p)
3068         char *names = pprint_rename(p->one->path, p->two->path);
3070         fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
3071         free(names);
3072         show_mode_change(file, p, 0);
3075 static void diff_summary(FILE *file, struct diff_filepair *p)
3077         switch(p->status) {
3078         case DIFF_STATUS_DELETED:
3079                 show_file_mode_name(file, "delete", p->one);
3080                 break;
3081         case DIFF_STATUS_ADDED:
3082                 show_file_mode_name(file, "create", p->two);
3083                 break;
3084         case DIFF_STATUS_COPIED:
3085                 show_rename_copy(file, "copy", p);
3086                 break;
3087         case DIFF_STATUS_RENAMED:
3088                 show_rename_copy(file, "rename", p);
3089                 break;
3090         default:
3091                 if (p->score) {
3092                         fputs(" rewrite ", file);
3093                         write_name_quoted(p->two->path, file, ' ');
3094                         fprintf(file, "(%d%%)\n", similarity_index(p));
3095                 }
3096                 show_mode_change(file, p, !p->score);
3097                 break;
3098         }
3101 struct patch_id_t {
3102         git_SHA_CTX *ctx;
3103         int patchlen;
3104 };
3106 static int remove_space(char *line, int len)
3108         int i;
3109         char *dst = line;
3110         unsigned char c;
3112         for (i = 0; i < len; i++)
3113                 if (!isspace((c = line[i])))
3114                         *dst++ = c;
3116         return dst - line;
3119 static void patch_id_consume(void *priv, char *line, unsigned long len)
3121         struct patch_id_t *data = priv;
3122         int new_len;
3124         /* Ignore line numbers when computing the SHA1 of the patch */
3125         if (!prefixcmp(line, "@@ -"))
3126                 return;
3128         new_len = remove_space(line, len);
3130         git_SHA1_Update(data->ctx, line, new_len);
3131         data->patchlen += new_len;
3134 /* returns 0 upon success, and writes result into sha1 */
3135 static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
3137         struct diff_queue_struct *q = &diff_queued_diff;
3138         int i;
3139         git_SHA_CTX ctx;
3140         struct patch_id_t data;
3141         char buffer[PATH_MAX * 4 + 20];
3143         git_SHA1_Init(&ctx);
3144         memset(&data, 0, sizeof(struct patch_id_t));
3145         data.ctx = &ctx;
3147         for (i = 0; i < q->nr; i++) {
3148                 xpparam_t xpp;
3149                 xdemitconf_t xecfg;
3150                 xdemitcb_t ecb;
3151                 mmfile_t mf1, mf2;
3152                 struct diff_filepair *p = q->queue[i];
3153                 int len1, len2;
3155                 memset(&xpp, 0, sizeof(xpp));
3156                 memset(&xecfg, 0, sizeof(xecfg));
3157                 if (p->status == 0)
3158                         return error("internal diff status error");
3159                 if (p->status == DIFF_STATUS_UNKNOWN)
3160                         continue;
3161                 if (diff_unmodified_pair(p))
3162                         continue;
3163                 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3164                     (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3165                         continue;
3166                 if (DIFF_PAIR_UNMERGED(p))
3167                         continue;
3169                 diff_fill_sha1_info(p->one);
3170                 diff_fill_sha1_info(p->two);
3171                 if (fill_mmfile(&mf1, p->one) < 0 ||
3172                                 fill_mmfile(&mf2, p->two) < 0)
3173                         return error("unable to read files to diff");
3175                 len1 = remove_space(p->one->path, strlen(p->one->path));
3176                 len2 = remove_space(p->two->path, strlen(p->two->path));
3177                 if (p->one->mode == 0)
3178                         len1 = snprintf(buffer, sizeof(buffer),
3179                                         "diff--gita/%.*sb/%.*s"
3180                                         "newfilemode%06o"
3181                                         "---/dev/null"
3182                                         "+++b/%.*s",
3183                                         len1, p->one->path,
3184                                         len2, p->two->path,
3185                                         p->two->mode,
3186                                         len2, p->two->path);
3187                 else if (p->two->mode == 0)
3188                         len1 = snprintf(buffer, sizeof(buffer),
3189                                         "diff--gita/%.*sb/%.*s"
3190                                         "deletedfilemode%06o"
3191                                         "---a/%.*s"
3192                                         "+++/dev/null",
3193                                         len1, p->one->path,
3194                                         len2, p->two->path,
3195                                         p->one->mode,
3196                                         len1, p->one->path);
3197                 else
3198                         len1 = snprintf(buffer, sizeof(buffer),
3199                                         "diff--gita/%.*sb/%.*s"
3200                                         "---a/%.*s"
3201                                         "+++b/%.*s",
3202                                         len1, p->one->path,
3203                                         len2, p->two->path,
3204                                         len1, p->one->path,
3205                                         len2, p->two->path);
3206                 git_SHA1_Update(&ctx, buffer, len1);
3208                 xpp.flags = XDF_NEED_MINIMAL;
3209                 xecfg.ctxlen = 3;
3210                 xecfg.flags = XDL_EMIT_FUNCNAMES;
3211                 xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
3212                               &xpp, &xecfg, &ecb);
3213         }
3215         git_SHA1_Final(sha1, &ctx);
3216         return 0;
3219 int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
3221         struct diff_queue_struct *q = &diff_queued_diff;
3222         int i;
3223         int result = diff_get_patch_id(options, sha1);
3225         for (i = 0; i < q->nr; i++)
3226                 diff_free_filepair(q->queue[i]);
3228         free(q->queue);
3229         q->queue = NULL;
3230         q->nr = q->alloc = 0;
3232         return result;
3235 static int is_summary_empty(const struct diff_queue_struct *q)
3237         int i;
3239         for (i = 0; i < q->nr; i++) {
3240                 const struct diff_filepair *p = q->queue[i];
3242                 switch (p->status) {
3243                 case DIFF_STATUS_DELETED:
3244                 case DIFF_STATUS_ADDED:
3245                 case DIFF_STATUS_COPIED:
3246                 case DIFF_STATUS_RENAMED:
3247                         return 0;
3248                 default:
3249                         if (p->score)
3250                                 return 0;
3251                         if (p->one->mode && p->two->mode &&
3252                             p->one->mode != p->two->mode)
3253                                 return 0;
3254                         break;
3255                 }
3256         }
3257         return 1;
3260 void diff_flush(struct diff_options *options)
3262         struct diff_queue_struct *q = &diff_queued_diff;
3263         int i, output_format = options->output_format;
3264         int separator = 0;
3266         /*
3267          * Order: raw, stat, summary, patch
3268          * or:    name/name-status/checkdiff (other bits clear)
3269          */
3270         if (!q->nr)
3271                 goto free_queue;
3273         if (output_format & (DIFF_FORMAT_RAW |
3274                              DIFF_FORMAT_NAME |
3275                              DIFF_FORMAT_NAME_STATUS |
3276                              DIFF_FORMAT_CHECKDIFF)) {
3277                 for (i = 0; i < q->nr; i++) {
3278                         struct diff_filepair *p = q->queue[i];
3279                         if (check_pair_status(p))
3280                                 flush_one_pair(p, options);
3281                 }
3282                 separator++;
3283         }
3285         if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT)) {
3286                 struct diffstat_t diffstat;
3288                 memset(&diffstat, 0, sizeof(struct diffstat_t));
3289                 for (i = 0; i < q->nr; i++) {
3290                         struct diff_filepair *p = q->queue[i];
3291                         if (check_pair_status(p))
3292                                 diff_flush_stat(p, options, &diffstat);
3293                 }
3294                 if (output_format & DIFF_FORMAT_NUMSTAT)
3295                         show_numstat(&diffstat, options);
3296                 if (output_format & DIFF_FORMAT_DIFFSTAT)
3297                         show_stats(&diffstat, options);
3298                 if (output_format & DIFF_FORMAT_SHORTSTAT)
3299                         show_shortstats(&diffstat, options);
3300                 free_diffstat_info(&diffstat);
3301                 separator++;
3302         }
3303         if (output_format & DIFF_FORMAT_DIRSTAT)
3304                 show_dirstat(options);
3306         if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
3307                 for (i = 0; i < q->nr; i++)
3308                         diff_summary(options->file, q->queue[i]);
3309                 separator++;
3310         }
3312         if (output_format & DIFF_FORMAT_PATCH) {
3313                 if (separator) {
3314                         putc(options->line_termination, options->file);
3315                         if (options->stat_sep) {
3316                                 /* attach patch instead of inline */
3317                                 fputs(options->stat_sep, options->file);
3318                         }
3319                 }
3321                 for (i = 0; i < q->nr; i++) {
3322                         struct diff_filepair *p = q->queue[i];
3323                         if (check_pair_status(p))
3324                                 diff_flush_patch(p, options);
3325                 }
3326         }
3328         if (output_format & DIFF_FORMAT_CALLBACK)
3329                 options->format_callback(q, options, options->format_callback_data);
3331         for (i = 0; i < q->nr; i++)
3332                 diff_free_filepair(q->queue[i]);
3333 free_queue:
3334         free(q->queue);
3335         q->queue = NULL;
3336         q->nr = q->alloc = 0;
3337         if (options->close_file)
3338                 fclose(options->file);
3341 static void diffcore_apply_filter(const char *filter)
3343         int i;
3344         struct diff_queue_struct *q = &diff_queued_diff;
3345         struct diff_queue_struct outq;
3346         outq.queue = NULL;
3347         outq.nr = outq.alloc = 0;
3349         if (!filter)
3350                 return;
3352         if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
3353                 int found;
3354                 for (i = found = 0; !found && i < q->nr; i++) {
3355                         struct diff_filepair *p = q->queue[i];
3356                         if (((p->status == DIFF_STATUS_MODIFIED) &&
3357                              ((p->score &&
3358                                strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
3359                               (!p->score &&
3360                                strchr(filter, DIFF_STATUS_MODIFIED)))) ||
3361                             ((p->status != DIFF_STATUS_MODIFIED) &&
3362                              strchr(filter, p->status)))
3363                                 found++;
3364                 }
3365                 if (found)
3366                         return;
3368                 /* otherwise we will clear the whole queue
3369                  * by copying the empty outq at the end of this
3370                  * function, but first clear the current entries
3371                  * in the queue.
3372                  */
3373                 for (i = 0; i < q->nr; i++)
3374                         diff_free_filepair(q->queue[i]);
3375         }
3376         else {
3377                 /* Only the matching ones */
3378                 for (i = 0; i < q->nr; i++) {
3379                         struct diff_filepair *p = q->queue[i];
3381                         if (((p->status == DIFF_STATUS_MODIFIED) &&
3382                              ((p->score &&
3383                                strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
3384                               (!p->score &&
3385                                strchr(filter, DIFF_STATUS_MODIFIED)))) ||
3386                             ((p->status != DIFF_STATUS_MODIFIED) &&
3387                              strchr(filter, p->status)))
3388                                 diff_q(&outq, p);
3389                         else
3390                                 diff_free_filepair(p);
3391                 }
3392         }
3393         free(q->queue);
3394         *q = outq;
3397 /* Check whether two filespecs with the same mode and size are identical */
3398 static int diff_filespec_is_identical(struct diff_filespec *one,
3399                                       struct diff_filespec *two)
3401         if (S_ISGITLINK(one->mode))
3402                 return 0;
3403         if (diff_populate_filespec(one, 0))
3404                 return 0;
3405         if (diff_populate_filespec(two, 0))
3406                 return 0;
3407         return !memcmp(one->data, two->data, one->size);
3410 static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
3412         int i;
3413         struct diff_queue_struct *q = &diff_queued_diff;
3414         struct diff_queue_struct outq;
3415         outq.queue = NULL;
3416         outq.nr = outq.alloc = 0;
3418         for (i = 0; i < q->nr; i++) {
3419                 struct diff_filepair *p = q->queue[i];
3421                 /*
3422                  * 1. Entries that come from stat info dirtyness
3423                  *    always have both sides (iow, not create/delete),
3424                  *    one side of the object name is unknown, with
3425                  *    the same mode and size.  Keep the ones that
3426                  *    do not match these criteria.  They have real
3427                  *    differences.
3428                  *
3429                  * 2. At this point, the file is known to be modified,
3430                  *    with the same mode and size, and the object
3431                  *    name of one side is unknown.  Need to inspect
3432                  *    the identical contents.
3433                  */
3434                 if (!DIFF_FILE_VALID(p->one) || /* (1) */
3435                     !DIFF_FILE_VALID(p->two) ||
3436                     (p->one->sha1_valid && p->two->sha1_valid) ||
3437                     (p->one->mode != p->two->mode) ||
3438                     diff_populate_filespec(p->one, 1) ||
3439                     diff_populate_filespec(p->two, 1) ||
3440                     (p->one->size != p->two->size) ||
3441                     !diff_filespec_is_identical(p->one, p->two)) /* (2) */
3442                         diff_q(&outq, p);
3443                 else {
3444                         /*
3445                          * The caller can subtract 1 from skip_stat_unmatch
3446                          * to determine how many paths were dirty only
3447                          * due to stat info mismatch.
3448                          */
3449                         if (!DIFF_OPT_TST(diffopt, NO_INDEX))
3450                                 diffopt->skip_stat_unmatch++;
3451                         diff_free_filepair(p);
3452                 }
3453         }
3454         free(q->queue);
3455         *q = outq;
3458 void diffcore_std(struct diff_options *options)
3460         if (options->skip_stat_unmatch)
3461                 diffcore_skip_stat_unmatch(options);
3462         if (options->break_opt != -1)
3463                 diffcore_break(options->break_opt);
3464         if (options->detect_rename)
3465                 diffcore_rename(options);
3466         if (options->break_opt != -1)
3467                 diffcore_merge_broken();
3468         if (options->pickaxe)
3469                 diffcore_pickaxe(options->pickaxe, options->pickaxe_opts);
3470         if (options->orderfile)
3471                 diffcore_order(options->orderfile);
3472         diff_resolve_rename_copy();
3473         diffcore_apply_filter(options->filter);
3475         if (diff_queued_diff.nr)
3476                 DIFF_OPT_SET(options, HAS_CHANGES);
3477         else
3478                 DIFF_OPT_CLR(options, HAS_CHANGES);
3481 int diff_result_code(struct diff_options *opt, int status)
3483         int result = 0;
3484         if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
3485             !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
3486                 return status;
3487         if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
3488             DIFF_OPT_TST(opt, HAS_CHANGES))
3489                 result |= 01;
3490         if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
3491             DIFF_OPT_TST(opt, CHECK_FAILED))
3492                 result |= 02;
3493         return result;
3496 void diff_addremove(struct diff_options *options,
3497                     int addremove, unsigned mode,
3498                     const unsigned char *sha1,
3499                     const char *concatpath)
3501         struct diff_filespec *one, *two;
3503         if (DIFF_OPT_TST(options, IGNORE_SUBMODULES) && S_ISGITLINK(mode))
3504                 return;
3506         /* This may look odd, but it is a preparation for
3507          * feeding "there are unchanged files which should
3508          * not produce diffs, but when you are doing copy
3509          * detection you would need them, so here they are"
3510          * entries to the diff-core.  They will be prefixed
3511          * with something like '=' or '*' (I haven't decided
3512          * which but should not make any difference).
3513          * Feeding the same new and old to diff_change()
3514          * also has the same effect.
3515          * Before the final output happens, they are pruned after
3516          * merged into rename/copy pairs as appropriate.
3517          */
3518         if (DIFF_OPT_TST(options, REVERSE_DIFF))
3519                 addremove = (addremove == '+' ? '-' :
3520                              addremove == '-' ? '+' : addremove);
3522         if (options->prefix &&
3523             strncmp(concatpath, options->prefix, options->prefix_length))
3524                 return;
3526         one = alloc_filespec(concatpath);
3527         two = alloc_filespec(concatpath);
3529         if (addremove != '+')
3530                 fill_filespec(one, sha1, mode);
3531         if (addremove != '-')
3532                 fill_filespec(two, sha1, mode);
3534         diff_queue(&diff_queued_diff, one, two);
3535         DIFF_OPT_SET(options, HAS_CHANGES);
3538 void diff_change(struct diff_options *options,
3539                  unsigned old_mode, unsigned new_mode,
3540                  const unsigned char *old_sha1,
3541                  const unsigned char *new_sha1,
3542                  const char *concatpath)
3544         struct diff_filespec *one, *two;
3546         if (DIFF_OPT_TST(options, IGNORE_SUBMODULES) && S_ISGITLINK(old_mode)
3547                         && S_ISGITLINK(new_mode))
3548                 return;
3550         if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
3551                 unsigned tmp;
3552                 const unsigned char *tmp_c;
3553                 tmp = old_mode; old_mode = new_mode; new_mode = tmp;
3554                 tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
3555         }
3557         if (options->prefix &&
3558             strncmp(concatpath, options->prefix, options->prefix_length))
3559                 return;
3561         one = alloc_filespec(concatpath);
3562         two = alloc_filespec(concatpath);
3563         fill_filespec(one, old_sha1, old_mode);
3564         fill_filespec(two, new_sha1, new_mode);
3566         diff_queue(&diff_queued_diff, one, two);
3567         DIFF_OPT_SET(options, HAS_CHANGES);
3570 void diff_unmerge(struct diff_options *options,
3571                   const char *path,
3572                   unsigned mode, const unsigned char *sha1)
3574         struct diff_filespec *one, *two;
3576         if (options->prefix &&
3577             strncmp(path, options->prefix, options->prefix_length))
3578                 return;
3580         one = alloc_filespec(path);
3581         two = alloc_filespec(path);
3582         fill_filespec(one, sha1, mode);
3583         diff_queue(&diff_queued_diff, one, two)->is_unmerged = 1;
3586 static char *run_textconv(const char *pgm, struct diff_filespec *spec,
3587                 size_t *outsize)
3589         struct diff_tempfile *temp;
3590         const char *argv[3];
3591         const char **arg = argv;
3592         struct child_process child;
3593         struct strbuf buf = STRBUF_INIT;
3595         temp = prepare_temp_file(spec->path, spec);
3596         *arg++ = pgm;
3597         *arg++ = temp->name;
3598         *arg = NULL;
3600         memset(&child, 0, sizeof(child));
3601         child.argv = argv;
3602         child.out = -1;
3603         if (start_command(&child) != 0 ||
3604             strbuf_read(&buf, child.out, 0) < 0 ||
3605             finish_command(&child) != 0) {
3606                 strbuf_release(&buf);
3607                 remove_tempfile();
3608                 error("error running textconv command '%s'", pgm);
3609                 return NULL;
3610         }
3611         remove_tempfile();
3613         return strbuf_detach(&buf, outsize);