Code

drop odd return value semantics from userdiff_config
[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"
16 #include "submodule.h"
17 #include "ll-merge.h"
19 #ifdef NO_FAST_WORKING_DIRECTORY
20 #define FAST_WORKING_DIRECTORY 0
21 #else
22 #define FAST_WORKING_DIRECTORY 1
23 #endif
25 static int diff_detect_rename_default;
26 static int diff_rename_limit_default = 400;
27 static int diff_suppress_blank_empty;
28 int diff_use_color_default = -1;
29 static const char *diff_word_regex_cfg;
30 static const char *external_diff_cmd_cfg;
31 int diff_auto_refresh_index = 1;
32 static int diff_mnemonic_prefix;
33 static int diff_no_prefix;
34 static int diff_dirstat_permille_default = 30;
35 static struct diff_options default_diff_options;
37 static char diff_colors[][COLOR_MAXLEN] = {
38         GIT_COLOR_RESET,
39         GIT_COLOR_NORMAL,       /* PLAIN */
40         GIT_COLOR_BOLD,         /* METAINFO */
41         GIT_COLOR_CYAN,         /* FRAGINFO */
42         GIT_COLOR_RED,          /* OLD */
43         GIT_COLOR_GREEN,        /* NEW */
44         GIT_COLOR_YELLOW,       /* COMMIT */
45         GIT_COLOR_BG_RED,       /* WHITESPACE */
46         GIT_COLOR_NORMAL,       /* FUNCINFO */
47 };
49 static int parse_diff_color_slot(const char *var, int ofs)
50 {
51         if (!strcasecmp(var+ofs, "plain"))
52                 return DIFF_PLAIN;
53         if (!strcasecmp(var+ofs, "meta"))
54                 return DIFF_METAINFO;
55         if (!strcasecmp(var+ofs, "frag"))
56                 return DIFF_FRAGINFO;
57         if (!strcasecmp(var+ofs, "old"))
58                 return DIFF_FILE_OLD;
59         if (!strcasecmp(var+ofs, "new"))
60                 return DIFF_FILE_NEW;
61         if (!strcasecmp(var+ofs, "commit"))
62                 return DIFF_COMMIT;
63         if (!strcasecmp(var+ofs, "whitespace"))
64                 return DIFF_WHITESPACE;
65         if (!strcasecmp(var+ofs, "func"))
66                 return DIFF_FUNCINFO;
67         return -1;
68 }
70 static int parse_dirstat_params(struct diff_options *options, const char *params,
71                                 struct strbuf *errmsg)
72 {
73         const char *p = params;
74         int p_len, ret = 0;
76         while (*p) {
77                 p_len = strchrnul(p, ',') - p;
78                 if (!memcmp(p, "changes", p_len)) {
79                         DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
80                         DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
81                 } else if (!memcmp(p, "lines", p_len)) {
82                         DIFF_OPT_SET(options, DIRSTAT_BY_LINE);
83                         DIFF_OPT_CLR(options, DIRSTAT_BY_FILE);
84                 } else if (!memcmp(p, "files", p_len)) {
85                         DIFF_OPT_CLR(options, DIRSTAT_BY_LINE);
86                         DIFF_OPT_SET(options, DIRSTAT_BY_FILE);
87                 } else if (!memcmp(p, "noncumulative", p_len)) {
88                         DIFF_OPT_CLR(options, DIRSTAT_CUMULATIVE);
89                 } else if (!memcmp(p, "cumulative", p_len)) {
90                         DIFF_OPT_SET(options, DIRSTAT_CUMULATIVE);
91                 } else if (isdigit(*p)) {
92                         char *end;
93                         int permille = strtoul(p, &end, 10) * 10;
94                         if (*end == '.' && isdigit(*++end)) {
95                                 /* only use first digit */
96                                 permille += *end - '0';
97                                 /* .. and ignore any further digits */
98                                 while (isdigit(*++end))
99                                         ; /* nothing */
100                         }
101                         if (end - p == p_len)
102                                 options->dirstat_permille = permille;
103                         else {
104                                 strbuf_addf(errmsg, _("  Failed to parse dirstat cut-off percentage '%.*s'\n"),
105                                             p_len, p);
106                                 ret++;
107                         }
108                 } else {
109                         strbuf_addf(errmsg, _("  Unknown dirstat parameter '%.*s'\n"),
110                                     p_len, p);
111                         ret++;
112                 }
114                 p += p_len;
116                 if (*p)
117                         p++; /* more parameters, swallow separator */
118         }
119         return ret;
122 static int git_config_rename(const char *var, const char *value)
124         if (!value)
125                 return DIFF_DETECT_RENAME;
126         if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
127                 return  DIFF_DETECT_COPY;
128         return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
131 /*
132  * These are to give UI layer defaults.
133  * The core-level commands such as git-diff-files should
134  * never be affected by the setting of diff.renames
135  * the user happens to have in the configuration file.
136  */
137 int git_diff_ui_config(const char *var, const char *value, void *cb)
139         if (!strcmp(var, "diff.color") || !strcmp(var, "color.diff")) {
140                 diff_use_color_default = git_config_colorbool(var, value);
141                 return 0;
142         }
143         if (!strcmp(var, "diff.renames")) {
144                 diff_detect_rename_default = git_config_rename(var, value);
145                 return 0;
146         }
147         if (!strcmp(var, "diff.autorefreshindex")) {
148                 diff_auto_refresh_index = git_config_bool(var, value);
149                 return 0;
150         }
151         if (!strcmp(var, "diff.mnemonicprefix")) {
152                 diff_mnemonic_prefix = git_config_bool(var, value);
153                 return 0;
154         }
155         if (!strcmp(var, "diff.noprefix")) {
156                 diff_no_prefix = git_config_bool(var, value);
157                 return 0;
158         }
159         if (!strcmp(var, "diff.external"))
160                 return git_config_string(&external_diff_cmd_cfg, var, value);
161         if (!strcmp(var, "diff.wordregex"))
162                 return git_config_string(&diff_word_regex_cfg, var, value);
164         if (!strcmp(var, "diff.ignoresubmodules"))
165                 handle_ignore_submodules_arg(&default_diff_options, value);
167         if (git_color_config(var, value, cb) < 0)
168                 return -1;
170         return git_diff_basic_config(var, value, cb);
173 int git_diff_basic_config(const char *var, const char *value, void *cb)
175         if (!strcmp(var, "diff.renamelimit")) {
176                 diff_rename_limit_default = git_config_int(var, value);
177                 return 0;
178         }
180         if (userdiff_config(var, value) < 0)
181                 return -1;
183         if (!prefixcmp(var, "diff.color.") || !prefixcmp(var, "color.diff.")) {
184                 int slot = parse_diff_color_slot(var, 11);
185                 if (slot < 0)
186                         return 0;
187                 if (!value)
188                         return config_error_nonbool(var);
189                 color_parse(value, var, diff_colors[slot]);
190                 return 0;
191         }
193         /* like GNU diff's --suppress-blank-empty option  */
194         if (!strcmp(var, "diff.suppressblankempty") ||
195                         /* for backwards compatibility */
196                         !strcmp(var, "diff.suppress-blank-empty")) {
197                 diff_suppress_blank_empty = git_config_bool(var, value);
198                 return 0;
199         }
201         if (!strcmp(var, "diff.dirstat")) {
202                 struct strbuf errmsg = STRBUF_INIT;
203                 default_diff_options.dirstat_permille = diff_dirstat_permille_default;
204                 if (parse_dirstat_params(&default_diff_options, value, &errmsg))
205                         warning(_("Found errors in 'diff.dirstat' config variable:\n%s"),
206                                 errmsg.buf);
207                 strbuf_release(&errmsg);
208                 diff_dirstat_permille_default = default_diff_options.dirstat_permille;
209                 return 0;
210         }
212         if (!prefixcmp(var, "submodule."))
213                 return parse_submodule_config_option(var, value);
215         return git_default_config(var, value, cb);
218 static char *quote_two(const char *one, const char *two)
220         int need_one = quote_c_style(one, NULL, NULL, 1);
221         int need_two = quote_c_style(two, NULL, NULL, 1);
222         struct strbuf res = STRBUF_INIT;
224         if (need_one + need_two) {
225                 strbuf_addch(&res, '"');
226                 quote_c_style(one, &res, NULL, 1);
227                 quote_c_style(two, &res, NULL, 1);
228                 strbuf_addch(&res, '"');
229         } else {
230                 strbuf_addstr(&res, one);
231                 strbuf_addstr(&res, two);
232         }
233         return strbuf_detach(&res, NULL);
236 static const char *external_diff(void)
238         static const char *external_diff_cmd = NULL;
239         static int done_preparing = 0;
241         if (done_preparing)
242                 return external_diff_cmd;
243         external_diff_cmd = getenv("GIT_EXTERNAL_DIFF");
244         if (!external_diff_cmd)
245                 external_diff_cmd = external_diff_cmd_cfg;
246         done_preparing = 1;
247         return external_diff_cmd;
250 static struct diff_tempfile {
251         const char *name; /* filename external diff should read from */
252         char hex[41];
253         char mode[10];
254         char tmp_path[PATH_MAX];
255 } diff_temp[2];
257 typedef unsigned long (*sane_truncate_fn)(char *line, unsigned long len);
259 struct emit_callback {
260         int color_diff;
261         unsigned ws_rule;
262         int blank_at_eof_in_preimage;
263         int blank_at_eof_in_postimage;
264         int lno_in_preimage;
265         int lno_in_postimage;
266         sane_truncate_fn truncate;
267         const char **label_path;
268         struct diff_words_data *diff_words;
269         struct diff_options *opt;
270         int *found_changesp;
271         struct strbuf *header;
272 };
274 static int count_lines(const char *data, int size)
276         int count, ch, completely_empty = 1, nl_just_seen = 0;
277         count = 0;
278         while (0 < size--) {
279                 ch = *data++;
280                 if (ch == '\n') {
281                         count++;
282                         nl_just_seen = 1;
283                         completely_empty = 0;
284                 }
285                 else {
286                         nl_just_seen = 0;
287                         completely_empty = 0;
288                 }
289         }
290         if (completely_empty)
291                 return 0;
292         if (!nl_just_seen)
293                 count++; /* no trailing newline */
294         return count;
297 static int fill_mmfile(mmfile_t *mf, struct diff_filespec *one)
299         if (!DIFF_FILE_VALID(one)) {
300                 mf->ptr = (char *)""; /* does not matter */
301                 mf->size = 0;
302                 return 0;
303         }
304         else if (diff_populate_filespec(one, 0))
305                 return -1;
307         mf->ptr = one->data;
308         mf->size = one->size;
309         return 0;
312 /* like fill_mmfile, but only for size, so we can avoid retrieving blob */
313 static unsigned long diff_filespec_size(struct diff_filespec *one)
315         if (!DIFF_FILE_VALID(one))
316                 return 0;
317         diff_populate_filespec(one, 1);
318         return one->size;
321 static int count_trailing_blank(mmfile_t *mf, unsigned ws_rule)
323         char *ptr = mf->ptr;
324         long size = mf->size;
325         int cnt = 0;
327         if (!size)
328                 return cnt;
329         ptr += size - 1; /* pointing at the very end */
330         if (*ptr != '\n')
331                 ; /* incomplete line */
332         else
333                 ptr--; /* skip the last LF */
334         while (mf->ptr < ptr) {
335                 char *prev_eol;
336                 for (prev_eol = ptr; mf->ptr <= prev_eol; prev_eol--)
337                         if (*prev_eol == '\n')
338                                 break;
339                 if (!ws_blank_line(prev_eol + 1, ptr - prev_eol, ws_rule))
340                         break;
341                 cnt++;
342                 ptr = prev_eol - 1;
343         }
344         return cnt;
347 static void check_blank_at_eof(mmfile_t *mf1, mmfile_t *mf2,
348                                struct emit_callback *ecbdata)
350         int l1, l2, at;
351         unsigned ws_rule = ecbdata->ws_rule;
352         l1 = count_trailing_blank(mf1, ws_rule);
353         l2 = count_trailing_blank(mf2, ws_rule);
354         if (l2 <= l1) {
355                 ecbdata->blank_at_eof_in_preimage = 0;
356                 ecbdata->blank_at_eof_in_postimage = 0;
357                 return;
358         }
359         at = count_lines(mf1->ptr, mf1->size);
360         ecbdata->blank_at_eof_in_preimage = (at - l1) + 1;
362         at = count_lines(mf2->ptr, mf2->size);
363         ecbdata->blank_at_eof_in_postimage = (at - l2) + 1;
366 static void emit_line_0(struct diff_options *o, const char *set, const char *reset,
367                         int first, const char *line, int len)
369         int has_trailing_newline, has_trailing_carriage_return;
370         int nofirst;
371         FILE *file = o->file;
373         if (o->output_prefix) {
374                 struct strbuf *msg = NULL;
375                 msg = o->output_prefix(o, o->output_prefix_data);
376                 assert(msg);
377                 fwrite(msg->buf, msg->len, 1, file);
378         }
380         if (len == 0) {
381                 has_trailing_newline = (first == '\n');
382                 has_trailing_carriage_return = (!has_trailing_newline &&
383                                                 (first == '\r'));
384                 nofirst = has_trailing_newline || has_trailing_carriage_return;
385         } else {
386                 has_trailing_newline = (len > 0 && line[len-1] == '\n');
387                 if (has_trailing_newline)
388                         len--;
389                 has_trailing_carriage_return = (len > 0 && line[len-1] == '\r');
390                 if (has_trailing_carriage_return)
391                         len--;
392                 nofirst = 0;
393         }
395         if (len || !nofirst) {
396                 fputs(set, file);
397                 if (!nofirst)
398                         fputc(first, file);
399                 fwrite(line, len, 1, file);
400                 fputs(reset, file);
401         }
402         if (has_trailing_carriage_return)
403                 fputc('\r', file);
404         if (has_trailing_newline)
405                 fputc('\n', file);
408 static void emit_line(struct diff_options *o, const char *set, const char *reset,
409                       const char *line, int len)
411         emit_line_0(o, set, reset, line[0], line+1, len-1);
414 static int new_blank_line_at_eof(struct emit_callback *ecbdata, const char *line, int len)
416         if (!((ecbdata->ws_rule & WS_BLANK_AT_EOF) &&
417               ecbdata->blank_at_eof_in_preimage &&
418               ecbdata->blank_at_eof_in_postimage &&
419               ecbdata->blank_at_eof_in_preimage <= ecbdata->lno_in_preimage &&
420               ecbdata->blank_at_eof_in_postimage <= ecbdata->lno_in_postimage))
421                 return 0;
422         return ws_blank_line(line, len, ecbdata->ws_rule);
425 static void emit_add_line(const char *reset,
426                           struct emit_callback *ecbdata,
427                           const char *line, int len)
429         const char *ws = diff_get_color(ecbdata->color_diff, DIFF_WHITESPACE);
430         const char *set = diff_get_color(ecbdata->color_diff, DIFF_FILE_NEW);
432         if (!*ws)
433                 emit_line_0(ecbdata->opt, set, reset, '+', line, len);
434         else if (new_blank_line_at_eof(ecbdata, line, len))
435                 /* Blank line at EOF - paint '+' as well */
436                 emit_line_0(ecbdata->opt, ws, reset, '+', line, len);
437         else {
438                 /* Emit just the prefix, then the rest. */
439                 emit_line_0(ecbdata->opt, set, reset, '+', "", 0);
440                 ws_check_emit(line, len, ecbdata->ws_rule,
441                               ecbdata->opt->file, set, reset, ws);
442         }
445 static void emit_hunk_header(struct emit_callback *ecbdata,
446                              const char *line, int len)
448         const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
449         const char *frag = diff_get_color(ecbdata->color_diff, DIFF_FRAGINFO);
450         const char *func = diff_get_color(ecbdata->color_diff, DIFF_FUNCINFO);
451         const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
452         static const char atat[2] = { '@', '@' };
453         const char *cp, *ep;
454         struct strbuf msgbuf = STRBUF_INIT;
455         int org_len = len;
456         int i = 1;
458         /*
459          * As a hunk header must begin with "@@ -<old>, +<new> @@",
460          * it always is at least 10 bytes long.
461          */
462         if (len < 10 ||
463             memcmp(line, atat, 2) ||
464             !(ep = memmem(line + 2, len - 2, atat, 2))) {
465                 emit_line(ecbdata->opt, plain, reset, line, len);
466                 return;
467         }
468         ep += 2; /* skip over @@ */
470         /* The hunk header in fraginfo color */
471         strbuf_add(&msgbuf, frag, strlen(frag));
472         strbuf_add(&msgbuf, line, ep - line);
473         strbuf_add(&msgbuf, reset, strlen(reset));
475         /*
476          * trailing "\r\n"
477          */
478         for ( ; i < 3; i++)
479                 if (line[len - i] == '\r' || line[len - i] == '\n')
480                         len--;
482         /* blank before the func header */
483         for (cp = ep; ep - line < len; ep++)
484                 if (*ep != ' ' && *ep != '\t')
485                         break;
486         if (ep != cp) {
487                 strbuf_add(&msgbuf, plain, strlen(plain));
488                 strbuf_add(&msgbuf, cp, ep - cp);
489                 strbuf_add(&msgbuf, reset, strlen(reset));
490         }
492         if (ep < line + len) {
493                 strbuf_add(&msgbuf, func, strlen(func));
494                 strbuf_add(&msgbuf, ep, line + len - ep);
495                 strbuf_add(&msgbuf, reset, strlen(reset));
496         }
498         strbuf_add(&msgbuf, line + len, org_len - len);
499         emit_line(ecbdata->opt, "", "", msgbuf.buf, msgbuf.len);
500         strbuf_release(&msgbuf);
503 static struct diff_tempfile *claim_diff_tempfile(void) {
504         int i;
505         for (i = 0; i < ARRAY_SIZE(diff_temp); i++)
506                 if (!diff_temp[i].name)
507                         return diff_temp + i;
508         die("BUG: diff is failing to clean up its tempfiles");
511 static int remove_tempfile_installed;
513 static void remove_tempfile(void)
515         int i;
516         for (i = 0; i < ARRAY_SIZE(diff_temp); i++) {
517                 if (diff_temp[i].name == diff_temp[i].tmp_path)
518                         unlink_or_warn(diff_temp[i].name);
519                 diff_temp[i].name = NULL;
520         }
523 static void remove_tempfile_on_signal(int signo)
525         remove_tempfile();
526         sigchain_pop(signo);
527         raise(signo);
530 static void print_line_count(FILE *file, int count)
532         switch (count) {
533         case 0:
534                 fprintf(file, "0,0");
535                 break;
536         case 1:
537                 fprintf(file, "1");
538                 break;
539         default:
540                 fprintf(file, "1,%d", count);
541                 break;
542         }
545 static void emit_rewrite_lines(struct emit_callback *ecb,
546                                int prefix, const char *data, int size)
548         const char *endp = NULL;
549         static const char *nneof = " No newline at end of file\n";
550         const char *old = diff_get_color(ecb->color_diff, DIFF_FILE_OLD);
551         const char *reset = diff_get_color(ecb->color_diff, DIFF_RESET);
553         while (0 < size) {
554                 int len;
556                 endp = memchr(data, '\n', size);
557                 len = endp ? (endp - data + 1) : size;
558                 if (prefix != '+') {
559                         ecb->lno_in_preimage++;
560                         emit_line_0(ecb->opt, old, reset, '-',
561                                     data, len);
562                 } else {
563                         ecb->lno_in_postimage++;
564                         emit_add_line(reset, ecb, data, len);
565                 }
566                 size -= len;
567                 data += len;
568         }
569         if (!endp) {
570                 const char *plain = diff_get_color(ecb->color_diff,
571                                                    DIFF_PLAIN);
572                 emit_line_0(ecb->opt, plain, reset, '\\',
573                             nneof, strlen(nneof));
574         }
577 static void emit_rewrite_diff(const char *name_a,
578                               const char *name_b,
579                               struct diff_filespec *one,
580                               struct diff_filespec *two,
581                               struct userdiff_driver *textconv_one,
582                               struct userdiff_driver *textconv_two,
583                               struct diff_options *o)
585         int lc_a, lc_b;
586         const char *name_a_tab, *name_b_tab;
587         const char *metainfo = diff_get_color(o->use_color, DIFF_METAINFO);
588         const char *fraginfo = diff_get_color(o->use_color, DIFF_FRAGINFO);
589         const char *reset = diff_get_color(o->use_color, DIFF_RESET);
590         static struct strbuf a_name = STRBUF_INIT, b_name = STRBUF_INIT;
591         const char *a_prefix, *b_prefix;
592         char *data_one, *data_two;
593         size_t size_one, size_two;
594         struct emit_callback ecbdata;
595         char *line_prefix = "";
596         struct strbuf *msgbuf;
598         if (o && o->output_prefix) {
599                 msgbuf = o->output_prefix(o, o->output_prefix_data);
600                 line_prefix = msgbuf->buf;
601         }
603         if (diff_mnemonic_prefix && DIFF_OPT_TST(o, REVERSE_DIFF)) {
604                 a_prefix = o->b_prefix;
605                 b_prefix = o->a_prefix;
606         } else {
607                 a_prefix = o->a_prefix;
608                 b_prefix = o->b_prefix;
609         }
611         name_a += (*name_a == '/');
612         name_b += (*name_b == '/');
613         name_a_tab = strchr(name_a, ' ') ? "\t" : "";
614         name_b_tab = strchr(name_b, ' ') ? "\t" : "";
616         strbuf_reset(&a_name);
617         strbuf_reset(&b_name);
618         quote_two_c_style(&a_name, a_prefix, name_a, 0);
619         quote_two_c_style(&b_name, b_prefix, name_b, 0);
621         size_one = fill_textconv(textconv_one, one, &data_one);
622         size_two = fill_textconv(textconv_two, two, &data_two);
624         memset(&ecbdata, 0, sizeof(ecbdata));
625         ecbdata.color_diff = want_color(o->use_color);
626         ecbdata.found_changesp = &o->found_changes;
627         ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
628         ecbdata.opt = o;
629         if (ecbdata.ws_rule & WS_BLANK_AT_EOF) {
630                 mmfile_t mf1, mf2;
631                 mf1.ptr = (char *)data_one;
632                 mf2.ptr = (char *)data_two;
633                 mf1.size = size_one;
634                 mf2.size = size_two;
635                 check_blank_at_eof(&mf1, &mf2, &ecbdata);
636         }
637         ecbdata.lno_in_preimage = 1;
638         ecbdata.lno_in_postimage = 1;
640         lc_a = count_lines(data_one, size_one);
641         lc_b = count_lines(data_two, size_two);
642         fprintf(o->file,
643                 "%s%s--- %s%s%s\n%s%s+++ %s%s%s\n%s%s@@ -",
644                 line_prefix, metainfo, a_name.buf, name_a_tab, reset,
645                 line_prefix, metainfo, b_name.buf, name_b_tab, reset,
646                 line_prefix, fraginfo);
647         if (!o->irreversible_delete)
648                 print_line_count(o->file, lc_a);
649         else
650                 fprintf(o->file, "?,?");
651         fprintf(o->file, " +");
652         print_line_count(o->file, lc_b);
653         fprintf(o->file, " @@%s\n", reset);
654         if (lc_a && !o->irreversible_delete)
655                 emit_rewrite_lines(&ecbdata, '-', data_one, size_one);
656         if (lc_b)
657                 emit_rewrite_lines(&ecbdata, '+', data_two, size_two);
658         if (textconv_one)
659                 free((char *)data_one);
660         if (textconv_two)
661                 free((char *)data_two);
664 struct diff_words_buffer {
665         mmfile_t text;
666         long alloc;
667         struct diff_words_orig {
668                 const char *begin, *end;
669         } *orig;
670         int orig_nr, orig_alloc;
671 };
673 static void diff_words_append(char *line, unsigned long len,
674                 struct diff_words_buffer *buffer)
676         ALLOC_GROW(buffer->text.ptr, buffer->text.size + len, buffer->alloc);
677         line++;
678         len--;
679         memcpy(buffer->text.ptr + buffer->text.size, line, len);
680         buffer->text.size += len;
681         buffer->text.ptr[buffer->text.size] = '\0';
684 struct diff_words_style_elem {
685         const char *prefix;
686         const char *suffix;
687         const char *color; /* NULL; filled in by the setup code if
688                             * color is enabled */
689 };
691 struct diff_words_style {
692         enum diff_words_type type;
693         struct diff_words_style_elem new, old, ctx;
694         const char *newline;
695 };
697 static struct diff_words_style diff_words_styles[] = {
698         { DIFF_WORDS_PORCELAIN, {"+", "\n"}, {"-", "\n"}, {" ", "\n"}, "~\n" },
699         { DIFF_WORDS_PLAIN, {"{+", "+}"}, {"[-", "-]"}, {"", ""}, "\n" },
700         { DIFF_WORDS_COLOR, {"", ""}, {"", ""}, {"", ""}, "\n" }
701 };
703 struct diff_words_data {
704         struct diff_words_buffer minus, plus;
705         const char *current_plus;
706         int last_minus;
707         struct diff_options *opt;
708         regex_t *word_regex;
709         enum diff_words_type type;
710         struct diff_words_style *style;
711 };
713 static int fn_out_diff_words_write_helper(FILE *fp,
714                                           struct diff_words_style_elem *st_el,
715                                           const char *newline,
716                                           size_t count, const char *buf,
717                                           const char *line_prefix)
719         int print = 0;
721         while (count) {
722                 char *p = memchr(buf, '\n', count);
723                 if (print)
724                         fputs(line_prefix, fp);
725                 if (p != buf) {
726                         if (st_el->color && fputs(st_el->color, fp) < 0)
727                                 return -1;
728                         if (fputs(st_el->prefix, fp) < 0 ||
729                             fwrite(buf, p ? p - buf : count, 1, fp) != 1 ||
730                             fputs(st_el->suffix, fp) < 0)
731                                 return -1;
732                         if (st_el->color && *st_el->color
733                             && fputs(GIT_COLOR_RESET, fp) < 0)
734                                 return -1;
735                 }
736                 if (!p)
737                         return 0;
738                 if (fputs(newline, fp) < 0)
739                         return -1;
740                 count -= p + 1 - buf;
741                 buf = p + 1;
742                 print = 1;
743         }
744         return 0;
747 /*
748  * '--color-words' algorithm can be described as:
749  *
750  *   1. collect a the minus/plus lines of a diff hunk, divided into
751  *      minus-lines and plus-lines;
752  *
753  *   2. break both minus-lines and plus-lines into words and
754  *      place them into two mmfile_t with one word for each line;
755  *
756  *   3. use xdiff to run diff on the two mmfile_t to get the words level diff;
757  *
758  * And for the common parts of the both file, we output the plus side text.
759  * diff_words->current_plus is used to trace the current position of the plus file
760  * which printed. diff_words->last_minus is used to trace the last minus word
761  * printed.
762  *
763  * For '--graph' to work with '--color-words', we need to output the graph prefix
764  * on each line of color words output. Generally, there are two conditions on
765  * which we should output the prefix.
766  *
767  *   1. diff_words->last_minus == 0 &&
768  *      diff_words->current_plus == diff_words->plus.text.ptr
769  *
770  *      that is: the plus text must start as a new line, and if there is no minus
771  *      word printed, a graph prefix must be printed.
772  *
773  *   2. diff_words->current_plus > diff_words->plus.text.ptr &&
774  *      *(diff_words->current_plus - 1) == '\n'
775  *
776  *      that is: a graph prefix must be printed following a '\n'
777  */
778 static int color_words_output_graph_prefix(struct diff_words_data *diff_words)
780         if ((diff_words->last_minus == 0 &&
781                 diff_words->current_plus == diff_words->plus.text.ptr) ||
782                 (diff_words->current_plus > diff_words->plus.text.ptr &&
783                 *(diff_words->current_plus - 1) == '\n')) {
784                 return 1;
785         } else {
786                 return 0;
787         }
790 static void fn_out_diff_words_aux(void *priv, char *line, unsigned long len)
792         struct diff_words_data *diff_words = priv;
793         struct diff_words_style *style = diff_words->style;
794         int minus_first, minus_len, plus_first, plus_len;
795         const char *minus_begin, *minus_end, *plus_begin, *plus_end;
796         struct diff_options *opt = diff_words->opt;
797         struct strbuf *msgbuf;
798         char *line_prefix = "";
800         if (line[0] != '@' || parse_hunk_header(line, len,
801                         &minus_first, &minus_len, &plus_first, &plus_len))
802                 return;
804         assert(opt);
805         if (opt->output_prefix) {
806                 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
807                 line_prefix = msgbuf->buf;
808         }
810         /* POSIX requires that first be decremented by one if len == 0... */
811         if (minus_len) {
812                 minus_begin = diff_words->minus.orig[minus_first].begin;
813                 minus_end =
814                         diff_words->minus.orig[minus_first + minus_len - 1].end;
815         } else
816                 minus_begin = minus_end =
817                         diff_words->minus.orig[minus_first].end;
819         if (plus_len) {
820                 plus_begin = diff_words->plus.orig[plus_first].begin;
821                 plus_end = diff_words->plus.orig[plus_first + plus_len - 1].end;
822         } else
823                 plus_begin = plus_end = diff_words->plus.orig[plus_first].end;
825         if (color_words_output_graph_prefix(diff_words)) {
826                 fputs(line_prefix, diff_words->opt->file);
827         }
828         if (diff_words->current_plus != plus_begin) {
829                 fn_out_diff_words_write_helper(diff_words->opt->file,
830                                 &style->ctx, style->newline,
831                                 plus_begin - diff_words->current_plus,
832                                 diff_words->current_plus, line_prefix);
833                 if (*(plus_begin - 1) == '\n')
834                         fputs(line_prefix, diff_words->opt->file);
835         }
836         if (minus_begin != minus_end) {
837                 fn_out_diff_words_write_helper(diff_words->opt->file,
838                                 &style->old, style->newline,
839                                 minus_end - minus_begin, minus_begin,
840                                 line_prefix);
841         }
842         if (plus_begin != plus_end) {
843                 fn_out_diff_words_write_helper(diff_words->opt->file,
844                                 &style->new, style->newline,
845                                 plus_end - plus_begin, plus_begin,
846                                 line_prefix);
847         }
849         diff_words->current_plus = plus_end;
850         diff_words->last_minus = minus_first;
853 /* This function starts looking at *begin, and returns 0 iff a word was found. */
854 static int find_word_boundaries(mmfile_t *buffer, regex_t *word_regex,
855                 int *begin, int *end)
857         if (word_regex && *begin < buffer->size) {
858                 regmatch_t match[1];
859                 if (!regexec(word_regex, buffer->ptr + *begin, 1, match, 0)) {
860                         char *p = memchr(buffer->ptr + *begin + match[0].rm_so,
861                                         '\n', match[0].rm_eo - match[0].rm_so);
862                         *end = p ? p - buffer->ptr : match[0].rm_eo + *begin;
863                         *begin += match[0].rm_so;
864                         return *begin >= *end;
865                 }
866                 return -1;
867         }
869         /* find the next word */
870         while (*begin < buffer->size && isspace(buffer->ptr[*begin]))
871                 (*begin)++;
872         if (*begin >= buffer->size)
873                 return -1;
875         /* find the end of the word */
876         *end = *begin + 1;
877         while (*end < buffer->size && !isspace(buffer->ptr[*end]))
878                 (*end)++;
880         return 0;
883 /*
884  * This function splits the words in buffer->text, stores the list with
885  * newline separator into out, and saves the offsets of the original words
886  * in buffer->orig.
887  */
888 static void diff_words_fill(struct diff_words_buffer *buffer, mmfile_t *out,
889                 regex_t *word_regex)
891         int i, j;
892         long alloc = 0;
894         out->size = 0;
895         out->ptr = NULL;
897         /* fake an empty "0th" word */
898         ALLOC_GROW(buffer->orig, 1, buffer->orig_alloc);
899         buffer->orig[0].begin = buffer->orig[0].end = buffer->text.ptr;
900         buffer->orig_nr = 1;
902         for (i = 0; i < buffer->text.size; i++) {
903                 if (find_word_boundaries(&buffer->text, word_regex, &i, &j))
904                         return;
906                 /* store original boundaries */
907                 ALLOC_GROW(buffer->orig, buffer->orig_nr + 1,
908                                 buffer->orig_alloc);
909                 buffer->orig[buffer->orig_nr].begin = buffer->text.ptr + i;
910                 buffer->orig[buffer->orig_nr].end = buffer->text.ptr + j;
911                 buffer->orig_nr++;
913                 /* store one word */
914                 ALLOC_GROW(out->ptr, out->size + j - i + 1, alloc);
915                 memcpy(out->ptr + out->size, buffer->text.ptr + i, j - i);
916                 out->ptr[out->size + j - i] = '\n';
917                 out->size += j - i + 1;
919                 i = j - 1;
920         }
923 /* this executes the word diff on the accumulated buffers */
924 static void diff_words_show(struct diff_words_data *diff_words)
926         xpparam_t xpp;
927         xdemitconf_t xecfg;
928         mmfile_t minus, plus;
929         struct diff_words_style *style = diff_words->style;
931         struct diff_options *opt = diff_words->opt;
932         struct strbuf *msgbuf;
933         char *line_prefix = "";
935         assert(opt);
936         if (opt->output_prefix) {
937                 msgbuf = opt->output_prefix(opt, opt->output_prefix_data);
938                 line_prefix = msgbuf->buf;
939         }
941         /* special case: only removal */
942         if (!diff_words->plus.text.size) {
943                 fputs(line_prefix, diff_words->opt->file);
944                 fn_out_diff_words_write_helper(diff_words->opt->file,
945                         &style->old, style->newline,
946                         diff_words->minus.text.size,
947                         diff_words->minus.text.ptr, line_prefix);
948                 diff_words->minus.text.size = 0;
949                 return;
950         }
952         diff_words->current_plus = diff_words->plus.text.ptr;
953         diff_words->last_minus = 0;
955         memset(&xpp, 0, sizeof(xpp));
956         memset(&xecfg, 0, sizeof(xecfg));
957         diff_words_fill(&diff_words->minus, &minus, diff_words->word_regex);
958         diff_words_fill(&diff_words->plus, &plus, diff_words->word_regex);
959         xpp.flags = 0;
960         /* as only the hunk header will be parsed, we need a 0-context */
961         xecfg.ctxlen = 0;
962         xdi_diff_outf(&minus, &plus, fn_out_diff_words_aux, diff_words,
963                       &xpp, &xecfg);
964         free(minus.ptr);
965         free(plus.ptr);
966         if (diff_words->current_plus != diff_words->plus.text.ptr +
967                         diff_words->plus.text.size) {
968                 if (color_words_output_graph_prefix(diff_words))
969                         fputs(line_prefix, diff_words->opt->file);
970                 fn_out_diff_words_write_helper(diff_words->opt->file,
971                         &style->ctx, style->newline,
972                         diff_words->plus.text.ptr + diff_words->plus.text.size
973                         - diff_words->current_plus, diff_words->current_plus,
974                         line_prefix);
975         }
976         diff_words->minus.text.size = diff_words->plus.text.size = 0;
979 /* In "color-words" mode, show word-diff of words accumulated in the buffer */
980 static void diff_words_flush(struct emit_callback *ecbdata)
982         if (ecbdata->diff_words->minus.text.size ||
983             ecbdata->diff_words->plus.text.size)
984                 diff_words_show(ecbdata->diff_words);
987 static void free_diff_words_data(struct emit_callback *ecbdata)
989         if (ecbdata->diff_words) {
990                 diff_words_flush(ecbdata);
991                 free (ecbdata->diff_words->minus.text.ptr);
992                 free (ecbdata->diff_words->minus.orig);
993                 free (ecbdata->diff_words->plus.text.ptr);
994                 free (ecbdata->diff_words->plus.orig);
995                 if (ecbdata->diff_words->word_regex) {
996                         regfree(ecbdata->diff_words->word_regex);
997                         free(ecbdata->diff_words->word_regex);
998                 }
999                 free(ecbdata->diff_words);
1000                 ecbdata->diff_words = NULL;
1001         }
1004 const char *diff_get_color(int diff_use_color, enum color_diff ix)
1006         if (want_color(diff_use_color))
1007                 return diff_colors[ix];
1008         return "";
1011 static unsigned long sane_truncate_line(struct emit_callback *ecb, char *line, unsigned long len)
1013         const char *cp;
1014         unsigned long allot;
1015         size_t l = len;
1017         if (ecb->truncate)
1018                 return ecb->truncate(line, len);
1019         cp = line;
1020         allot = l;
1021         while (0 < l) {
1022                 (void) utf8_width(&cp, &l);
1023                 if (!cp)
1024                         break; /* truncated in the middle? */
1025         }
1026         return allot - l;
1029 static void find_lno(const char *line, struct emit_callback *ecbdata)
1031         const char *p;
1032         ecbdata->lno_in_preimage = 0;
1033         ecbdata->lno_in_postimage = 0;
1034         p = strchr(line, '-');
1035         if (!p)
1036                 return; /* cannot happen */
1037         ecbdata->lno_in_preimage = strtol(p + 1, NULL, 10);
1038         p = strchr(p, '+');
1039         if (!p)
1040                 return; /* cannot happen */
1041         ecbdata->lno_in_postimage = strtol(p + 1, NULL, 10);
1044 static void fn_out_consume(void *priv, char *line, unsigned long len)
1046         struct emit_callback *ecbdata = priv;
1047         const char *meta = diff_get_color(ecbdata->color_diff, DIFF_METAINFO);
1048         const char *plain = diff_get_color(ecbdata->color_diff, DIFF_PLAIN);
1049         const char *reset = diff_get_color(ecbdata->color_diff, DIFF_RESET);
1050         struct diff_options *o = ecbdata->opt;
1051         char *line_prefix = "";
1052         struct strbuf *msgbuf;
1054         if (o && o->output_prefix) {
1055                 msgbuf = o->output_prefix(o, o->output_prefix_data);
1056                 line_prefix = msgbuf->buf;
1057         }
1059         if (ecbdata->header) {
1060                 fprintf(ecbdata->opt->file, "%s", ecbdata->header->buf);
1061                 strbuf_reset(ecbdata->header);
1062                 ecbdata->header = NULL;
1063         }
1064         *(ecbdata->found_changesp) = 1;
1066         if (ecbdata->label_path[0]) {
1067                 const char *name_a_tab, *name_b_tab;
1069                 name_a_tab = strchr(ecbdata->label_path[0], ' ') ? "\t" : "";
1070                 name_b_tab = strchr(ecbdata->label_path[1], ' ') ? "\t" : "";
1072                 fprintf(ecbdata->opt->file, "%s%s--- %s%s%s\n",
1073                         line_prefix, meta, ecbdata->label_path[0], reset, name_a_tab);
1074                 fprintf(ecbdata->opt->file, "%s%s+++ %s%s%s\n",
1075                         line_prefix, meta, ecbdata->label_path[1], reset, name_b_tab);
1076                 ecbdata->label_path[0] = ecbdata->label_path[1] = NULL;
1077         }
1079         if (diff_suppress_blank_empty
1080             && len == 2 && line[0] == ' ' && line[1] == '\n') {
1081                 line[0] = '\n';
1082                 len = 1;
1083         }
1085         if (line[0] == '@') {
1086                 if (ecbdata->diff_words)
1087                         diff_words_flush(ecbdata);
1088                 len = sane_truncate_line(ecbdata, line, len);
1089                 find_lno(line, ecbdata);
1090                 emit_hunk_header(ecbdata, line, len);
1091                 if (line[len-1] != '\n')
1092                         putc('\n', ecbdata->opt->file);
1093                 return;
1094         }
1096         if (len < 1) {
1097                 emit_line(ecbdata->opt, reset, reset, line, len);
1098                 if (ecbdata->diff_words
1099                     && ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN)
1100                         fputs("~\n", ecbdata->opt->file);
1101                 return;
1102         }
1104         if (ecbdata->diff_words) {
1105                 if (line[0] == '-') {
1106                         diff_words_append(line, len,
1107                                           &ecbdata->diff_words->minus);
1108                         return;
1109                 } else if (line[0] == '+') {
1110                         diff_words_append(line, len,
1111                                           &ecbdata->diff_words->plus);
1112                         return;
1113                 } else if (!prefixcmp(line, "\\ ")) {
1114                         /*
1115                          * Eat the "no newline at eof" marker as if we
1116                          * saw a "+" or "-" line with nothing on it,
1117                          * and return without diff_words_flush() to
1118                          * defer processing. If this is the end of
1119                          * preimage, more "+" lines may come after it.
1120                          */
1121                         return;
1122                 }
1123                 diff_words_flush(ecbdata);
1124                 if (ecbdata->diff_words->type == DIFF_WORDS_PORCELAIN) {
1125                         emit_line(ecbdata->opt, plain, reset, line, len);
1126                         fputs("~\n", ecbdata->opt->file);
1127                 } else {
1128                         /*
1129                          * Skip the prefix character, if any.  With
1130                          * diff_suppress_blank_empty, there may be
1131                          * none.
1132                          */
1133                         if (line[0] != '\n') {
1134                               line++;
1135                               len--;
1136                         }
1137                         emit_line(ecbdata->opt, plain, reset, line, len);
1138                 }
1139                 return;
1140         }
1142         if (line[0] != '+') {
1143                 const char *color =
1144                         diff_get_color(ecbdata->color_diff,
1145                                        line[0] == '-' ? DIFF_FILE_OLD : DIFF_PLAIN);
1146                 ecbdata->lno_in_preimage++;
1147                 if (line[0] == ' ')
1148                         ecbdata->lno_in_postimage++;
1149                 emit_line(ecbdata->opt, color, reset, line, len);
1150         } else {
1151                 ecbdata->lno_in_postimage++;
1152                 emit_add_line(reset, ecbdata, line + 1, len - 1);
1153         }
1156 static char *pprint_rename(const char *a, const char *b)
1158         const char *old = a;
1159         const char *new = b;
1160         struct strbuf name = STRBUF_INIT;
1161         int pfx_length, sfx_length;
1162         int len_a = strlen(a);
1163         int len_b = strlen(b);
1164         int a_midlen, b_midlen;
1165         int qlen_a = quote_c_style(a, NULL, NULL, 0);
1166         int qlen_b = quote_c_style(b, NULL, NULL, 0);
1168         if (qlen_a || qlen_b) {
1169                 quote_c_style(a, &name, NULL, 0);
1170                 strbuf_addstr(&name, " => ");
1171                 quote_c_style(b, &name, NULL, 0);
1172                 return strbuf_detach(&name, NULL);
1173         }
1175         /* Find common prefix */
1176         pfx_length = 0;
1177         while (*old && *new && *old == *new) {
1178                 if (*old == '/')
1179                         pfx_length = old - a + 1;
1180                 old++;
1181                 new++;
1182         }
1184         /* Find common suffix */
1185         old = a + len_a;
1186         new = b + len_b;
1187         sfx_length = 0;
1188         while (a <= old && b <= new && *old == *new) {
1189                 if (*old == '/')
1190                         sfx_length = len_a - (old - a);
1191                 old--;
1192                 new--;
1193         }
1195         /*
1196          * pfx{mid-a => mid-b}sfx
1197          * {pfx-a => pfx-b}sfx
1198          * pfx{sfx-a => sfx-b}
1199          * name-a => name-b
1200          */
1201         a_midlen = len_a - pfx_length - sfx_length;
1202         b_midlen = len_b - pfx_length - sfx_length;
1203         if (a_midlen < 0)
1204                 a_midlen = 0;
1205         if (b_midlen < 0)
1206                 b_midlen = 0;
1208         strbuf_grow(&name, pfx_length + a_midlen + b_midlen + sfx_length + 7);
1209         if (pfx_length + sfx_length) {
1210                 strbuf_add(&name, a, pfx_length);
1211                 strbuf_addch(&name, '{');
1212         }
1213         strbuf_add(&name, a + pfx_length, a_midlen);
1214         strbuf_addstr(&name, " => ");
1215         strbuf_add(&name, b + pfx_length, b_midlen);
1216         if (pfx_length + sfx_length) {
1217                 strbuf_addch(&name, '}');
1218                 strbuf_add(&name, a + len_a - sfx_length, sfx_length);
1219         }
1220         return strbuf_detach(&name, NULL);
1223 struct diffstat_t {
1224         int nr;
1225         int alloc;
1226         struct diffstat_file {
1227                 char *from_name;
1228                 char *name;
1229                 char *print_name;
1230                 unsigned is_unmerged:1;
1231                 unsigned is_binary:1;
1232                 unsigned is_renamed:1;
1233                 uintmax_t added, deleted;
1234         } **files;
1235 };
1237 static struct diffstat_file *diffstat_add(struct diffstat_t *diffstat,
1238                                           const char *name_a,
1239                                           const char *name_b)
1241         struct diffstat_file *x;
1242         x = xcalloc(sizeof (*x), 1);
1243         if (diffstat->nr == diffstat->alloc) {
1244                 diffstat->alloc = alloc_nr(diffstat->alloc);
1245                 diffstat->files = xrealloc(diffstat->files,
1246                                 diffstat->alloc * sizeof(x));
1247         }
1248         diffstat->files[diffstat->nr++] = x;
1249         if (name_b) {
1250                 x->from_name = xstrdup(name_a);
1251                 x->name = xstrdup(name_b);
1252                 x->is_renamed = 1;
1253         }
1254         else {
1255                 x->from_name = NULL;
1256                 x->name = xstrdup(name_a);
1257         }
1258         return x;
1261 static void diffstat_consume(void *priv, char *line, unsigned long len)
1263         struct diffstat_t *diffstat = priv;
1264         struct diffstat_file *x = diffstat->files[diffstat->nr - 1];
1266         if (line[0] == '+')
1267                 x->added++;
1268         else if (line[0] == '-')
1269                 x->deleted++;
1272 const char mime_boundary_leader[] = "------------";
1274 static int scale_linear(int it, int width, int max_change)
1276         /*
1277          * make sure that at least one '-' is printed if there were deletions,
1278          * and likewise for '+'.
1279          */
1280         if (max_change < 2)
1281                 return it;
1282         return ((it - 1) * (width - 1) + max_change - 1) / (max_change - 1);
1285 static void show_name(FILE *file,
1286                       const char *prefix, const char *name, int len)
1288         fprintf(file, " %s%-*s |", prefix, len, name);
1291 static void show_graph(FILE *file, char ch, int cnt, const char *set, const char *reset)
1293         if (cnt <= 0)
1294                 return;
1295         fprintf(file, "%s", set);
1296         while (cnt--)
1297                 putc(ch, file);
1298         fprintf(file, "%s", reset);
1301 static void fill_print_name(struct diffstat_file *file)
1303         char *pname;
1305         if (file->print_name)
1306                 return;
1308         if (!file->is_renamed) {
1309                 struct strbuf buf = STRBUF_INIT;
1310                 if (quote_c_style(file->name, &buf, NULL, 0)) {
1311                         pname = strbuf_detach(&buf, NULL);
1312                 } else {
1313                         pname = file->name;
1314                         strbuf_release(&buf);
1315                 }
1316         } else {
1317                 pname = pprint_rename(file->from_name, file->name);
1318         }
1319         file->print_name = pname;
1322 static void show_stats(struct diffstat_t *data, struct diff_options *options)
1324         int i, len, add, del, adds = 0, dels = 0;
1325         uintmax_t max_change = 0, max_len = 0;
1326         int total_files = data->nr;
1327         int width, name_width, count;
1328         const char *reset, *add_c, *del_c;
1329         const char *line_prefix = "";
1330         int extra_shown = 0;
1331         struct strbuf *msg = NULL;
1333         if (data->nr == 0)
1334                 return;
1336         if (options->output_prefix) {
1337                 msg = options->output_prefix(options, options->output_prefix_data);
1338                 line_prefix = msg->buf;
1339         }
1341         width = options->stat_width ? options->stat_width : 80;
1342         name_width = options->stat_name_width ? options->stat_name_width : 50;
1343         count = options->stat_count ? options->stat_count : data->nr;
1345         /* Sanity: give at least 5 columns to the graph,
1346          * but leave at least 10 columns for the name.
1347          */
1348         if (width < 25)
1349                 width = 25;
1350         if (name_width < 10)
1351                 name_width = 10;
1352         else if (width < name_width + 15)
1353                 name_width = width - 15;
1355         /* Find the longest filename and max number of changes */
1356         reset = diff_get_color_opt(options, DIFF_RESET);
1357         add_c = diff_get_color_opt(options, DIFF_FILE_NEW);
1358         del_c = diff_get_color_opt(options, DIFF_FILE_OLD);
1360         for (i = 0; (i < count) && (i < data->nr); i++) {
1361                 struct diffstat_file *file = data->files[i];
1362                 uintmax_t change = file->added + file->deleted;
1363                 if (!data->files[i]->is_renamed &&
1364                          (change == 0)) {
1365                         count++; /* not shown == room for one more */
1366                         continue;
1367                 }
1368                 fill_print_name(file);
1369                 len = strlen(file->print_name);
1370                 if (max_len < len)
1371                         max_len = len;
1373                 if (file->is_binary || file->is_unmerged)
1374                         continue;
1375                 if (max_change < change)
1376                         max_change = change;
1377         }
1378         count = i; /* min(count, data->nr) */
1380         /* Compute the width of the graph part;
1381          * 10 is for one blank at the beginning of the line plus
1382          * " | count " between the name and the graph.
1383          *
1384          * From here on, name_width is the width of the name area,
1385          * and width is the width of the graph area.
1386          */
1387         name_width = (name_width < max_len) ? name_width : max_len;
1388         if (width < (name_width + 10) + max_change)
1389                 width = width - (name_width + 10);
1390         else
1391                 width = max_change;
1393         for (i = 0; i < count; i++) {
1394                 const char *prefix = "";
1395                 char *name = data->files[i]->print_name;
1396                 uintmax_t added = data->files[i]->added;
1397                 uintmax_t deleted = data->files[i]->deleted;
1398                 int name_len;
1400                 if (!data->files[i]->is_renamed &&
1401                          (added + deleted == 0)) {
1402                         total_files--;
1403                         continue;
1404                 }
1405                 /*
1406                  * "scale" the filename
1407                  */
1408                 len = name_width;
1409                 name_len = strlen(name);
1410                 if (name_width < name_len) {
1411                         char *slash;
1412                         prefix = "...";
1413                         len -= 3;
1414                         name += name_len - len;
1415                         slash = strchr(name, '/');
1416                         if (slash)
1417                                 name = slash;
1418                 }
1420                 if (data->files[i]->is_binary) {
1421                         fprintf(options->file, "%s", line_prefix);
1422                         show_name(options->file, prefix, name, len);
1423                         fprintf(options->file, "  Bin ");
1424                         fprintf(options->file, "%s%"PRIuMAX"%s",
1425                                 del_c, deleted, reset);
1426                         fprintf(options->file, " -> ");
1427                         fprintf(options->file, "%s%"PRIuMAX"%s",
1428                                 add_c, added, reset);
1429                         fprintf(options->file, " bytes");
1430                         fprintf(options->file, "\n");
1431                         continue;
1432                 }
1433                 else if (data->files[i]->is_unmerged) {
1434                         fprintf(options->file, "%s", line_prefix);
1435                         show_name(options->file, prefix, name, len);
1436                         fprintf(options->file, "  Unmerged\n");
1437                         continue;
1438                 }
1440                 /*
1441                  * scale the add/delete
1442                  */
1443                 add = added;
1444                 del = deleted;
1445                 adds += add;
1446                 dels += del;
1448                 if (width <= max_change) {
1449                         add = scale_linear(add, width, max_change);
1450                         del = scale_linear(del, width, max_change);
1451                 }
1452                 fprintf(options->file, "%s", line_prefix);
1453                 show_name(options->file, prefix, name, len);
1454                 fprintf(options->file, "%5"PRIuMAX"%s", added + deleted,
1455                                 added + deleted ? " " : "");
1456                 show_graph(options->file, '+', add, add_c, reset);
1457                 show_graph(options->file, '-', del, del_c, reset);
1458                 fprintf(options->file, "\n");
1459         }
1460         for (i = count; i < data->nr; i++) {
1461                 uintmax_t added = data->files[i]->added;
1462                 uintmax_t deleted = data->files[i]->deleted;
1463                 if (!data->files[i]->is_renamed &&
1464                          (added + deleted == 0)) {
1465                         total_files--;
1466                         continue;
1467                 }
1468                 adds += added;
1469                 dels += deleted;
1470                 if (!extra_shown)
1471                         fprintf(options->file, "%s ...\n", line_prefix);
1472                 extra_shown = 1;
1473         }
1474         fprintf(options->file, "%s", line_prefix);
1475         fprintf(options->file,
1476                " %d files changed, %d insertions(+), %d deletions(-)\n",
1477                total_files, adds, dels);
1480 static void show_shortstats(struct diffstat_t *data, struct diff_options *options)
1482         int i, adds = 0, dels = 0, total_files = data->nr;
1484         if (data->nr == 0)
1485                 return;
1487         for (i = 0; i < data->nr; i++) {
1488                 if (!data->files[i]->is_binary &&
1489                     !data->files[i]->is_unmerged) {
1490                         int added = data->files[i]->added;
1491                         int deleted= data->files[i]->deleted;
1492                         if (!data->files[i]->is_renamed &&
1493                             (added + deleted == 0)) {
1494                                 total_files--;
1495                         } else {
1496                                 adds += added;
1497                                 dels += deleted;
1498                         }
1499                 }
1500         }
1501         if (options->output_prefix) {
1502                 struct strbuf *msg = NULL;
1503                 msg = options->output_prefix(options,
1504                                 options->output_prefix_data);
1505                 fprintf(options->file, "%s", msg->buf);
1506         }
1507         fprintf(options->file, " %d files changed, %d insertions(+), %d deletions(-)\n",
1508                total_files, adds, dels);
1511 static void show_numstat(struct diffstat_t *data, struct diff_options *options)
1513         int i;
1515         if (data->nr == 0)
1516                 return;
1518         for (i = 0; i < data->nr; i++) {
1519                 struct diffstat_file *file = data->files[i];
1521                 if (options->output_prefix) {
1522                         struct strbuf *msg = NULL;
1523                         msg = options->output_prefix(options,
1524                                         options->output_prefix_data);
1525                         fprintf(options->file, "%s", msg->buf);
1526                 }
1528                 if (file->is_binary)
1529                         fprintf(options->file, "-\t-\t");
1530                 else
1531                         fprintf(options->file,
1532                                 "%"PRIuMAX"\t%"PRIuMAX"\t",
1533                                 file->added, file->deleted);
1534                 if (options->line_termination) {
1535                         fill_print_name(file);
1536                         if (!file->is_renamed)
1537                                 write_name_quoted(file->name, options->file,
1538                                                   options->line_termination);
1539                         else {
1540                                 fputs(file->print_name, options->file);
1541                                 putc(options->line_termination, options->file);
1542                         }
1543                 } else {
1544                         if (file->is_renamed) {
1545                                 putc('\0', options->file);
1546                                 write_name_quoted(file->from_name, options->file, '\0');
1547                         }
1548                         write_name_quoted(file->name, options->file, '\0');
1549                 }
1550         }
1553 struct dirstat_file {
1554         const char *name;
1555         unsigned long changed;
1556 };
1558 struct dirstat_dir {
1559         struct dirstat_file *files;
1560         int alloc, nr, permille, cumulative;
1561 };
1563 static long gather_dirstat(struct diff_options *opt, struct dirstat_dir *dir,
1564                 unsigned long changed, const char *base, int baselen)
1566         unsigned long this_dir = 0;
1567         unsigned int sources = 0;
1568         const char *line_prefix = "";
1569         struct strbuf *msg = NULL;
1571         if (opt->output_prefix) {
1572                 msg = opt->output_prefix(opt, opt->output_prefix_data);
1573                 line_prefix = msg->buf;
1574         }
1576         while (dir->nr) {
1577                 struct dirstat_file *f = dir->files;
1578                 int namelen = strlen(f->name);
1579                 unsigned long this;
1580                 char *slash;
1582                 if (namelen < baselen)
1583                         break;
1584                 if (memcmp(f->name, base, baselen))
1585                         break;
1586                 slash = strchr(f->name + baselen, '/');
1587                 if (slash) {
1588                         int newbaselen = slash + 1 - f->name;
1589                         this = gather_dirstat(opt, dir, changed, f->name, newbaselen);
1590                         sources++;
1591                 } else {
1592                         this = f->changed;
1593                         dir->files++;
1594                         dir->nr--;
1595                         sources += 2;
1596                 }
1597                 this_dir += this;
1598         }
1600         /*
1601          * We don't report dirstat's for
1602          *  - the top level
1603          *  - or cases where everything came from a single directory
1604          *    under this directory (sources == 1).
1605          */
1606         if (baselen && sources != 1) {
1607                 if (this_dir) {
1608                         int permille = this_dir * 1000 / changed;
1609                         if (permille >= dir->permille) {
1610                                 fprintf(opt->file, "%s%4d.%01d%% %.*s\n", line_prefix,
1611                                         permille / 10, permille % 10, baselen, base);
1612                                 if (!dir->cumulative)
1613                                         return 0;
1614                         }
1615                 }
1616         }
1617         return this_dir;
1620 static int dirstat_compare(const void *_a, const void *_b)
1622         const struct dirstat_file *a = _a;
1623         const struct dirstat_file *b = _b;
1624         return strcmp(a->name, b->name);
1627 static void show_dirstat(struct diff_options *options)
1629         int i;
1630         unsigned long changed;
1631         struct dirstat_dir dir;
1632         struct diff_queue_struct *q = &diff_queued_diff;
1634         dir.files = NULL;
1635         dir.alloc = 0;
1636         dir.nr = 0;
1637         dir.permille = options->dirstat_permille;
1638         dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1640         changed = 0;
1641         for (i = 0; i < q->nr; i++) {
1642                 struct diff_filepair *p = q->queue[i];
1643                 const char *name;
1644                 unsigned long copied, added, damage;
1645                 int content_changed;
1647                 name = p->two->path ? p->two->path : p->one->path;
1649                 if (p->one->sha1_valid && p->two->sha1_valid)
1650                         content_changed = hashcmp(p->one->sha1, p->two->sha1);
1651                 else
1652                         content_changed = 1;
1654                 if (!content_changed) {
1655                         /*
1656                          * The SHA1 has not changed, so pre-/post-content is
1657                          * identical. We can therefore skip looking at the
1658                          * file contents altogether.
1659                          */
1660                         damage = 0;
1661                         goto found_damage;
1662                 }
1664                 if (DIFF_OPT_TST(options, DIRSTAT_BY_FILE)) {
1665                         /*
1666                          * In --dirstat-by-file mode, we don't really need to
1667                          * look at the actual file contents at all.
1668                          * The fact that the SHA1 changed is enough for us to
1669                          * add this file to the list of results
1670                          * (with each file contributing equal damage).
1671                          */
1672                         damage = 1;
1673                         goto found_damage;
1674                 }
1676                 if (DIFF_FILE_VALID(p->one) && DIFF_FILE_VALID(p->two)) {
1677                         diff_populate_filespec(p->one, 0);
1678                         diff_populate_filespec(p->two, 0);
1679                         diffcore_count_changes(p->one, p->two, NULL, NULL, 0,
1680                                                &copied, &added);
1681                         diff_free_filespec_data(p->one);
1682                         diff_free_filespec_data(p->two);
1683                 } else if (DIFF_FILE_VALID(p->one)) {
1684                         diff_populate_filespec(p->one, 1);
1685                         copied = added = 0;
1686                         diff_free_filespec_data(p->one);
1687                 } else if (DIFF_FILE_VALID(p->two)) {
1688                         diff_populate_filespec(p->two, 1);
1689                         copied = 0;
1690                         added = p->two->size;
1691                         diff_free_filespec_data(p->two);
1692                 } else
1693                         continue;
1695                 /*
1696                  * Original minus copied is the removed material,
1697                  * added is the new material.  They are both damages
1698                  * made to the preimage.
1699                  * If the resulting damage is zero, we know that
1700                  * diffcore_count_changes() considers the two entries to
1701                  * be identical, but since content_changed is true, we
1702                  * know that there must have been _some_ kind of change,
1703                  * so we force all entries to have damage > 0.
1704                  */
1705                 damage = (p->one->size - copied) + added;
1706                 if (!damage)
1707                         damage = 1;
1709 found_damage:
1710                 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1711                 dir.files[dir.nr].name = name;
1712                 dir.files[dir.nr].changed = damage;
1713                 changed += damage;
1714                 dir.nr++;
1715         }
1717         /* This can happen even with many files, if everything was renames */
1718         if (!changed)
1719                 return;
1721         /* Show all directories with more than x% of the changes */
1722         qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1723         gather_dirstat(options, &dir, changed, "", 0);
1726 static void show_dirstat_by_line(struct diffstat_t *data, struct diff_options *options)
1728         int i;
1729         unsigned long changed;
1730         struct dirstat_dir dir;
1732         if (data->nr == 0)
1733                 return;
1735         dir.files = NULL;
1736         dir.alloc = 0;
1737         dir.nr = 0;
1738         dir.permille = options->dirstat_permille;
1739         dir.cumulative = DIFF_OPT_TST(options, DIRSTAT_CUMULATIVE);
1741         changed = 0;
1742         for (i = 0; i < data->nr; i++) {
1743                 struct diffstat_file *file = data->files[i];
1744                 unsigned long damage = file->added + file->deleted;
1745                 if (file->is_binary)
1746                         /*
1747                          * binary files counts bytes, not lines. Must find some
1748                          * way to normalize binary bytes vs. textual lines.
1749                          * The following heuristic assumes that there are 64
1750                          * bytes per "line".
1751                          * This is stupid and ugly, but very cheap...
1752                          */
1753                         damage = (damage + 63) / 64;
1754                 ALLOC_GROW(dir.files, dir.nr + 1, dir.alloc);
1755                 dir.files[dir.nr].name = file->name;
1756                 dir.files[dir.nr].changed = damage;
1757                 changed += damage;
1758                 dir.nr++;
1759         }
1761         /* This can happen even with many files, if everything was renames */
1762         if (!changed)
1763                 return;
1765         /* Show all directories with more than x% of the changes */
1766         qsort(dir.files, dir.nr, sizeof(dir.files[0]), dirstat_compare);
1767         gather_dirstat(options, &dir, changed, "", 0);
1770 static void free_diffstat_info(struct diffstat_t *diffstat)
1772         int i;
1773         for (i = 0; i < diffstat->nr; i++) {
1774                 struct diffstat_file *f = diffstat->files[i];
1775                 if (f->name != f->print_name)
1776                         free(f->print_name);
1777                 free(f->name);
1778                 free(f->from_name);
1779                 free(f);
1780         }
1781         free(diffstat->files);
1784 struct checkdiff_t {
1785         const char *filename;
1786         int lineno;
1787         int conflict_marker_size;
1788         struct diff_options *o;
1789         unsigned ws_rule;
1790         unsigned status;
1791 };
1793 static int is_conflict_marker(const char *line, int marker_size, unsigned long len)
1795         char firstchar;
1796         int cnt;
1798         if (len < marker_size + 1)
1799                 return 0;
1800         firstchar = line[0];
1801         switch (firstchar) {
1802         case '=': case '>': case '<': case '|':
1803                 break;
1804         default:
1805                 return 0;
1806         }
1807         for (cnt = 1; cnt < marker_size; cnt++)
1808                 if (line[cnt] != firstchar)
1809                         return 0;
1810         /* line[1] thru line[marker_size-1] are same as firstchar */
1811         if (len < marker_size + 1 || !isspace(line[marker_size]))
1812                 return 0;
1813         return 1;
1816 static void checkdiff_consume(void *priv, char *line, unsigned long len)
1818         struct checkdiff_t *data = priv;
1819         int marker_size = data->conflict_marker_size;
1820         const char *ws = diff_get_color(data->o->use_color, DIFF_WHITESPACE);
1821         const char *reset = diff_get_color(data->o->use_color, DIFF_RESET);
1822         const char *set = diff_get_color(data->o->use_color, DIFF_FILE_NEW);
1823         char *err;
1824         char *line_prefix = "";
1825         struct strbuf *msgbuf;
1827         assert(data->o);
1828         if (data->o->output_prefix) {
1829                 msgbuf = data->o->output_prefix(data->o,
1830                         data->o->output_prefix_data);
1831                 line_prefix = msgbuf->buf;
1832         }
1834         if (line[0] == '+') {
1835                 unsigned bad;
1836                 data->lineno++;
1837                 if (is_conflict_marker(line + 1, marker_size, len - 1)) {
1838                         data->status |= 1;
1839                         fprintf(data->o->file,
1840                                 "%s%s:%d: leftover conflict marker\n",
1841                                 line_prefix, data->filename, data->lineno);
1842                 }
1843                 bad = ws_check(line + 1, len - 1, data->ws_rule);
1844                 if (!bad)
1845                         return;
1846                 data->status |= bad;
1847                 err = whitespace_error_string(bad);
1848                 fprintf(data->o->file, "%s%s:%d: %s.\n",
1849                         line_prefix, data->filename, data->lineno, err);
1850                 free(err);
1851                 emit_line(data->o, set, reset, line, 1);
1852                 ws_check_emit(line + 1, len - 1, data->ws_rule,
1853                               data->o->file, set, reset, ws);
1854         } else if (line[0] == ' ') {
1855                 data->lineno++;
1856         } else if (line[0] == '@') {
1857                 char *plus = strchr(line, '+');
1858                 if (plus)
1859                         data->lineno = strtol(plus, NULL, 10) - 1;
1860                 else
1861                         die("invalid diff");
1862         }
1865 static unsigned char *deflate_it(char *data,
1866                                  unsigned long size,
1867                                  unsigned long *result_size)
1869         int bound;
1870         unsigned char *deflated;
1871         git_zstream stream;
1873         memset(&stream, 0, sizeof(stream));
1874         git_deflate_init(&stream, zlib_compression_level);
1875         bound = git_deflate_bound(&stream, size);
1876         deflated = xmalloc(bound);
1877         stream.next_out = deflated;
1878         stream.avail_out = bound;
1880         stream.next_in = (unsigned char *)data;
1881         stream.avail_in = size;
1882         while (git_deflate(&stream, Z_FINISH) == Z_OK)
1883                 ; /* nothing */
1884         git_deflate_end(&stream);
1885         *result_size = stream.total_out;
1886         return deflated;
1889 static void emit_binary_diff_body(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
1891         void *cp;
1892         void *delta;
1893         void *deflated;
1894         void *data;
1895         unsigned long orig_size;
1896         unsigned long delta_size;
1897         unsigned long deflate_size;
1898         unsigned long data_size;
1900         /* We could do deflated delta, or we could do just deflated two,
1901          * whichever is smaller.
1902          */
1903         delta = NULL;
1904         deflated = deflate_it(two->ptr, two->size, &deflate_size);
1905         if (one->size && two->size) {
1906                 delta = diff_delta(one->ptr, one->size,
1907                                    two->ptr, two->size,
1908                                    &delta_size, deflate_size);
1909                 if (delta) {
1910                         void *to_free = delta;
1911                         orig_size = delta_size;
1912                         delta = deflate_it(delta, delta_size, &delta_size);
1913                         free(to_free);
1914                 }
1915         }
1917         if (delta && delta_size < deflate_size) {
1918                 fprintf(file, "%sdelta %lu\n", prefix, orig_size);
1919                 free(deflated);
1920                 data = delta;
1921                 data_size = delta_size;
1922         }
1923         else {
1924                 fprintf(file, "%sliteral %lu\n", prefix, two->size);
1925                 free(delta);
1926                 data = deflated;
1927                 data_size = deflate_size;
1928         }
1930         /* emit data encoded in base85 */
1931         cp = data;
1932         while (data_size) {
1933                 int bytes = (52 < data_size) ? 52 : data_size;
1934                 char line[70];
1935                 data_size -= bytes;
1936                 if (bytes <= 26)
1937                         line[0] = bytes + 'A' - 1;
1938                 else
1939                         line[0] = bytes - 26 + 'a' - 1;
1940                 encode_85(line + 1, cp, bytes);
1941                 cp = (char *) cp + bytes;
1942                 fprintf(file, "%s", prefix);
1943                 fputs(line, file);
1944                 fputc('\n', file);
1945         }
1946         fprintf(file, "%s\n", prefix);
1947         free(data);
1950 static void emit_binary_diff(FILE *file, mmfile_t *one, mmfile_t *two, char *prefix)
1952         fprintf(file, "%sGIT binary patch\n", prefix);
1953         emit_binary_diff_body(file, one, two, prefix);
1954         emit_binary_diff_body(file, two, one, prefix);
1957 static void diff_filespec_load_driver(struct diff_filespec *one)
1959         /* Use already-loaded driver */
1960         if (one->driver)
1961                 return;
1963         if (S_ISREG(one->mode))
1964                 one->driver = userdiff_find_by_path(one->path);
1966         /* Fallback to default settings */
1967         if (!one->driver)
1968                 one->driver = userdiff_find_by_name("default");
1971 int diff_filespec_is_binary(struct diff_filespec *one)
1973         if (one->is_binary == -1) {
1974                 diff_filespec_load_driver(one);
1975                 if (one->driver->binary != -1)
1976                         one->is_binary = one->driver->binary;
1977                 else {
1978                         if (!one->data && DIFF_FILE_VALID(one))
1979                                 diff_populate_filespec(one, 0);
1980                         if (one->data)
1981                                 one->is_binary = buffer_is_binary(one->data,
1982                                                 one->size);
1983                         if (one->is_binary == -1)
1984                                 one->is_binary = 0;
1985                 }
1986         }
1987         return one->is_binary;
1990 static const struct userdiff_funcname *diff_funcname_pattern(struct diff_filespec *one)
1992         diff_filespec_load_driver(one);
1993         return one->driver->funcname.pattern ? &one->driver->funcname : NULL;
1996 static const char *userdiff_word_regex(struct diff_filespec *one)
1998         diff_filespec_load_driver(one);
1999         return one->driver->word_regex;
2002 void diff_set_mnemonic_prefix(struct diff_options *options, const char *a, const char *b)
2004         if (!options->a_prefix)
2005                 options->a_prefix = a;
2006         if (!options->b_prefix)
2007                 options->b_prefix = b;
2010 struct userdiff_driver *get_textconv(struct diff_filespec *one)
2012         if (!DIFF_FILE_VALID(one))
2013                 return NULL;
2015         diff_filespec_load_driver(one);
2016         return userdiff_get_textconv(one->driver);
2019 static void builtin_diff(const char *name_a,
2020                          const char *name_b,
2021                          struct diff_filespec *one,
2022                          struct diff_filespec *two,
2023                          const char *xfrm_msg,
2024                          int must_show_header,
2025                          struct diff_options *o,
2026                          int complete_rewrite)
2028         mmfile_t mf1, mf2;
2029         const char *lbl[2];
2030         char *a_one, *b_two;
2031         const char *set = diff_get_color_opt(o, DIFF_METAINFO);
2032         const char *reset = diff_get_color_opt(o, DIFF_RESET);
2033         const char *a_prefix, *b_prefix;
2034         struct userdiff_driver *textconv_one = NULL;
2035         struct userdiff_driver *textconv_two = NULL;
2036         struct strbuf header = STRBUF_INIT;
2037         struct strbuf *msgbuf;
2038         char *line_prefix = "";
2040         if (o->output_prefix) {
2041                 msgbuf = o->output_prefix(o, o->output_prefix_data);
2042                 line_prefix = msgbuf->buf;
2043         }
2045         if (DIFF_OPT_TST(o, SUBMODULE_LOG) &&
2046                         (!one->mode || S_ISGITLINK(one->mode)) &&
2047                         (!two->mode || S_ISGITLINK(two->mode))) {
2048                 const char *del = diff_get_color_opt(o, DIFF_FILE_OLD);
2049                 const char *add = diff_get_color_opt(o, DIFF_FILE_NEW);
2050                 show_submodule_summary(o->file, one ? one->path : two->path,
2051                                 one->sha1, two->sha1, two->dirty_submodule,
2052                                 del, add, reset);
2053                 return;
2054         }
2056         if (DIFF_OPT_TST(o, ALLOW_TEXTCONV)) {
2057                 textconv_one = get_textconv(one);
2058                 textconv_two = get_textconv(two);
2059         }
2061         diff_set_mnemonic_prefix(o, "a/", "b/");
2062         if (DIFF_OPT_TST(o, REVERSE_DIFF)) {
2063                 a_prefix = o->b_prefix;
2064                 b_prefix = o->a_prefix;
2065         } else {
2066                 a_prefix = o->a_prefix;
2067                 b_prefix = o->b_prefix;
2068         }
2070         /* Never use a non-valid filename anywhere if at all possible */
2071         name_a = DIFF_FILE_VALID(one) ? name_a : name_b;
2072         name_b = DIFF_FILE_VALID(two) ? name_b : name_a;
2074         a_one = quote_two(a_prefix, name_a + (*name_a == '/'));
2075         b_two = quote_two(b_prefix, name_b + (*name_b == '/'));
2076         lbl[0] = DIFF_FILE_VALID(one) ? a_one : "/dev/null";
2077         lbl[1] = DIFF_FILE_VALID(two) ? b_two : "/dev/null";
2078         strbuf_addf(&header, "%s%sdiff --git %s %s%s\n", line_prefix, set, a_one, b_two, reset);
2079         if (lbl[0][0] == '/') {
2080                 /* /dev/null */
2081                 strbuf_addf(&header, "%s%snew file mode %06o%s\n", line_prefix, set, two->mode, reset);
2082                 if (xfrm_msg)
2083                         strbuf_addstr(&header, xfrm_msg);
2084                 must_show_header = 1;
2085         }
2086         else if (lbl[1][0] == '/') {
2087                 strbuf_addf(&header, "%s%sdeleted file mode %06o%s\n", line_prefix, set, one->mode, reset);
2088                 if (xfrm_msg)
2089                         strbuf_addstr(&header, xfrm_msg);
2090                 must_show_header = 1;
2091         }
2092         else {
2093                 if (one->mode != two->mode) {
2094                         strbuf_addf(&header, "%s%sold mode %06o%s\n", line_prefix, set, one->mode, reset);
2095                         strbuf_addf(&header, "%s%snew mode %06o%s\n", line_prefix, set, two->mode, reset);
2096                         must_show_header = 1;
2097                 }
2098                 if (xfrm_msg)
2099                         strbuf_addstr(&header, xfrm_msg);
2101                 /*
2102                  * we do not run diff between different kind
2103                  * of objects.
2104                  */
2105                 if ((one->mode ^ two->mode) & S_IFMT)
2106                         goto free_ab_and_return;
2107                 if (complete_rewrite &&
2108                     (textconv_one || !diff_filespec_is_binary(one)) &&
2109                     (textconv_two || !diff_filespec_is_binary(two))) {
2110                         fprintf(o->file, "%s", header.buf);
2111                         strbuf_reset(&header);
2112                         emit_rewrite_diff(name_a, name_b, one, two,
2113                                                 textconv_one, textconv_two, o);
2114                         o->found_changes = 1;
2115                         goto free_ab_and_return;
2116                 }
2117         }
2119         if (o->irreversible_delete && lbl[1][0] == '/') {
2120                 fprintf(o->file, "%s", header.buf);
2121                 strbuf_reset(&header);
2122                 goto free_ab_and_return;
2123         } else if (!DIFF_OPT_TST(o, TEXT) &&
2124             ( (!textconv_one && diff_filespec_is_binary(one)) ||
2125               (!textconv_two && diff_filespec_is_binary(two)) )) {
2126                 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2127                         die("unable to read files to diff");
2128                 /* Quite common confusing case */
2129                 if (mf1.size == mf2.size &&
2130                     !memcmp(mf1.ptr, mf2.ptr, mf1.size)) {
2131                         if (must_show_header)
2132                                 fprintf(o->file, "%s", header.buf);
2133                         goto free_ab_and_return;
2134                 }
2135                 fprintf(o->file, "%s", header.buf);
2136                 strbuf_reset(&header);
2137                 if (DIFF_OPT_TST(o, BINARY))
2138                         emit_binary_diff(o->file, &mf1, &mf2, line_prefix);
2139                 else
2140                         fprintf(o->file, "%sBinary files %s and %s differ\n",
2141                                 line_prefix, lbl[0], lbl[1]);
2142                 o->found_changes = 1;
2143         } else {
2144                 /* Crazy xdl interfaces.. */
2145                 const char *diffopts = getenv("GIT_DIFF_OPTS");
2146                 xpparam_t xpp;
2147                 xdemitconf_t xecfg;
2148                 struct emit_callback ecbdata;
2149                 const struct userdiff_funcname *pe;
2151                 if (!DIFF_XDL_TST(o, WHITESPACE_FLAGS) || must_show_header) {
2152                         fprintf(o->file, "%s", header.buf);
2153                         strbuf_reset(&header);
2154                 }
2156                 mf1.size = fill_textconv(textconv_one, one, &mf1.ptr);
2157                 mf2.size = fill_textconv(textconv_two, two, &mf2.ptr);
2159                 pe = diff_funcname_pattern(one);
2160                 if (!pe)
2161                         pe = diff_funcname_pattern(two);
2163                 memset(&xpp, 0, sizeof(xpp));
2164                 memset(&xecfg, 0, sizeof(xecfg));
2165                 memset(&ecbdata, 0, sizeof(ecbdata));
2166                 ecbdata.label_path = lbl;
2167                 ecbdata.color_diff = want_color(o->use_color);
2168                 ecbdata.found_changesp = &o->found_changes;
2169                 ecbdata.ws_rule = whitespace_rule(name_b ? name_b : name_a);
2170                 if (ecbdata.ws_rule & WS_BLANK_AT_EOF)
2171                         check_blank_at_eof(&mf1, &mf2, &ecbdata);
2172                 ecbdata.opt = o;
2173                 ecbdata.header = header.len ? &header : NULL;
2174                 xpp.flags = o->xdl_opts;
2175                 xecfg.ctxlen = o->context;
2176                 xecfg.interhunkctxlen = o->interhunkcontext;
2177                 xecfg.flags = XDL_EMIT_FUNCNAMES;
2178                 if (DIFF_OPT_TST(o, FUNCCONTEXT))
2179                         xecfg.flags |= XDL_EMIT_FUNCCONTEXT;
2180                 if (pe)
2181                         xdiff_set_find_func(&xecfg, pe->pattern, pe->cflags);
2182                 if (!diffopts)
2183                         ;
2184                 else if (!prefixcmp(diffopts, "--unified="))
2185                         xecfg.ctxlen = strtoul(diffopts + 10, NULL, 10);
2186                 else if (!prefixcmp(diffopts, "-u"))
2187                         xecfg.ctxlen = strtoul(diffopts + 2, NULL, 10);
2188                 if (o->word_diff) {
2189                         int i;
2191                         ecbdata.diff_words =
2192                                 xcalloc(1, sizeof(struct diff_words_data));
2193                         ecbdata.diff_words->type = o->word_diff;
2194                         ecbdata.diff_words->opt = o;
2195                         if (!o->word_regex)
2196                                 o->word_regex = userdiff_word_regex(one);
2197                         if (!o->word_regex)
2198                                 o->word_regex = userdiff_word_regex(two);
2199                         if (!o->word_regex)
2200                                 o->word_regex = diff_word_regex_cfg;
2201                         if (o->word_regex) {
2202                                 ecbdata.diff_words->word_regex = (regex_t *)
2203                                         xmalloc(sizeof(regex_t));
2204                                 if (regcomp(ecbdata.diff_words->word_regex,
2205                                                 o->word_regex,
2206                                                 REG_EXTENDED | REG_NEWLINE))
2207                                         die ("Invalid regular expression: %s",
2208                                                         o->word_regex);
2209                         }
2210                         for (i = 0; i < ARRAY_SIZE(diff_words_styles); i++) {
2211                                 if (o->word_diff == diff_words_styles[i].type) {
2212                                         ecbdata.diff_words->style =
2213                                                 &diff_words_styles[i];
2214                                         break;
2215                                 }
2216                         }
2217                         if (want_color(o->use_color)) {
2218                                 struct diff_words_style *st = ecbdata.diff_words->style;
2219                                 st->old.color = diff_get_color_opt(o, DIFF_FILE_OLD);
2220                                 st->new.color = diff_get_color_opt(o, DIFF_FILE_NEW);
2221                                 st->ctx.color = diff_get_color_opt(o, DIFF_PLAIN);
2222                         }
2223                 }
2224                 xdi_diff_outf(&mf1, &mf2, fn_out_consume, &ecbdata,
2225                               &xpp, &xecfg);
2226                 if (o->word_diff)
2227                         free_diff_words_data(&ecbdata);
2228                 if (textconv_one)
2229                         free(mf1.ptr);
2230                 if (textconv_two)
2231                         free(mf2.ptr);
2232                 xdiff_clear_find_func(&xecfg);
2233         }
2235  free_ab_and_return:
2236         strbuf_release(&header);
2237         diff_free_filespec_data(one);
2238         diff_free_filespec_data(two);
2239         free(a_one);
2240         free(b_two);
2241         return;
2244 static void builtin_diffstat(const char *name_a, const char *name_b,
2245                              struct diff_filespec *one,
2246                              struct diff_filespec *two,
2247                              struct diffstat_t *diffstat,
2248                              struct diff_options *o,
2249                              int complete_rewrite)
2251         mmfile_t mf1, mf2;
2252         struct diffstat_file *data;
2254         data = diffstat_add(diffstat, name_a, name_b);
2256         if (!one || !two) {
2257                 data->is_unmerged = 1;
2258                 return;
2259         }
2261         if (diff_filespec_is_binary(one) || diff_filespec_is_binary(two)) {
2262                 data->is_binary = 1;
2263                 data->added = diff_filespec_size(two);
2264                 data->deleted = diff_filespec_size(one);
2265         }
2267         else if (complete_rewrite) {
2268                 diff_populate_filespec(one, 0);
2269                 diff_populate_filespec(two, 0);
2270                 data->deleted = count_lines(one->data, one->size);
2271                 data->added = count_lines(two->data, two->size);
2272         }
2274         else {
2275                 /* Crazy xdl interfaces.. */
2276                 xpparam_t xpp;
2277                 xdemitconf_t xecfg;
2279                 if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2280                         die("unable to read files to diff");
2282                 memset(&xpp, 0, sizeof(xpp));
2283                 memset(&xecfg, 0, sizeof(xecfg));
2284                 xpp.flags = o->xdl_opts;
2285                 xecfg.ctxlen = o->context;
2286                 xecfg.interhunkctxlen = o->interhunkcontext;
2287                 xdi_diff_outf(&mf1, &mf2, diffstat_consume, diffstat,
2288                               &xpp, &xecfg);
2289         }
2291         diff_free_filespec_data(one);
2292         diff_free_filespec_data(two);
2295 static void builtin_checkdiff(const char *name_a, const char *name_b,
2296                               const char *attr_path,
2297                               struct diff_filespec *one,
2298                               struct diff_filespec *two,
2299                               struct diff_options *o)
2301         mmfile_t mf1, mf2;
2302         struct checkdiff_t data;
2304         if (!two)
2305                 return;
2307         memset(&data, 0, sizeof(data));
2308         data.filename = name_b ? name_b : name_a;
2309         data.lineno = 0;
2310         data.o = o;
2311         data.ws_rule = whitespace_rule(attr_path);
2312         data.conflict_marker_size = ll_merge_marker_size(attr_path);
2314         if (fill_mmfile(&mf1, one) < 0 || fill_mmfile(&mf2, two) < 0)
2315                 die("unable to read files to diff");
2317         /*
2318          * All the other codepaths check both sides, but not checking
2319          * the "old" side here is deliberate.  We are checking the newly
2320          * introduced changes, and as long as the "new" side is text, we
2321          * can and should check what it introduces.
2322          */
2323         if (diff_filespec_is_binary(two))
2324                 goto free_and_return;
2325         else {
2326                 /* Crazy xdl interfaces.. */
2327                 xpparam_t xpp;
2328                 xdemitconf_t xecfg;
2330                 memset(&xpp, 0, sizeof(xpp));
2331                 memset(&xecfg, 0, sizeof(xecfg));
2332                 xecfg.ctxlen = 1; /* at least one context line */
2333                 xpp.flags = 0;
2334                 xdi_diff_outf(&mf1, &mf2, checkdiff_consume, &data,
2335                               &xpp, &xecfg);
2337                 if (data.ws_rule & WS_BLANK_AT_EOF) {
2338                         struct emit_callback ecbdata;
2339                         int blank_at_eof;
2341                         ecbdata.ws_rule = data.ws_rule;
2342                         check_blank_at_eof(&mf1, &mf2, &ecbdata);
2343                         blank_at_eof = ecbdata.blank_at_eof_in_postimage;
2345                         if (blank_at_eof) {
2346                                 static char *err;
2347                                 if (!err)
2348                                         err = whitespace_error_string(WS_BLANK_AT_EOF);
2349                                 fprintf(o->file, "%s:%d: %s.\n",
2350                                         data.filename, blank_at_eof, err);
2351                                 data.status = 1; /* report errors */
2352                         }
2353                 }
2354         }
2355  free_and_return:
2356         diff_free_filespec_data(one);
2357         diff_free_filespec_data(two);
2358         if (data.status)
2359                 DIFF_OPT_SET(o, CHECK_FAILED);
2362 struct diff_filespec *alloc_filespec(const char *path)
2364         int namelen = strlen(path);
2365         struct diff_filespec *spec = xmalloc(sizeof(*spec) + namelen + 1);
2367         memset(spec, 0, sizeof(*spec));
2368         spec->path = (char *)(spec + 1);
2369         memcpy(spec->path, path, namelen+1);
2370         spec->count = 1;
2371         spec->is_binary = -1;
2372         return spec;
2375 void free_filespec(struct diff_filespec *spec)
2377         if (!--spec->count) {
2378                 diff_free_filespec_data(spec);
2379                 free(spec);
2380         }
2383 void fill_filespec(struct diff_filespec *spec, const unsigned char *sha1,
2384                    unsigned short mode)
2386         if (mode) {
2387                 spec->mode = canon_mode(mode);
2388                 hashcpy(spec->sha1, sha1);
2389                 spec->sha1_valid = !is_null_sha1(sha1);
2390         }
2393 /*
2394  * Given a name and sha1 pair, if the index tells us the file in
2395  * the work tree has that object contents, return true, so that
2396  * prepare_temp_file() does not have to inflate and extract.
2397  */
2398 static int reuse_worktree_file(const char *name, const unsigned char *sha1, int want_file)
2400         struct cache_entry *ce;
2401         struct stat st;
2402         int pos, len;
2404         /*
2405          * We do not read the cache ourselves here, because the
2406          * benchmark with my previous version that always reads cache
2407          * shows that it makes things worse for diff-tree comparing
2408          * two linux-2.6 kernel trees in an already checked out work
2409          * tree.  This is because most diff-tree comparisons deal with
2410          * only a small number of files, while reading the cache is
2411          * expensive for a large project, and its cost outweighs the
2412          * savings we get by not inflating the object to a temporary
2413          * file.  Practically, this code only helps when we are used
2414          * by diff-cache --cached, which does read the cache before
2415          * calling us.
2416          */
2417         if (!active_cache)
2418                 return 0;
2420         /* We want to avoid the working directory if our caller
2421          * doesn't need the data in a normal file, this system
2422          * is rather slow with its stat/open/mmap/close syscalls,
2423          * and the object is contained in a pack file.  The pack
2424          * is probably already open and will be faster to obtain
2425          * the data through than the working directory.  Loose
2426          * objects however would tend to be slower as they need
2427          * to be individually opened and inflated.
2428          */
2429         if (!FAST_WORKING_DIRECTORY && !want_file && has_sha1_pack(sha1))
2430                 return 0;
2432         len = strlen(name);
2433         pos = cache_name_pos(name, len);
2434         if (pos < 0)
2435                 return 0;
2436         ce = active_cache[pos];
2438         /*
2439          * This is not the sha1 we are looking for, or
2440          * unreusable because it is not a regular file.
2441          */
2442         if (hashcmp(sha1, ce->sha1) || !S_ISREG(ce->ce_mode))
2443                 return 0;
2445         /*
2446          * If ce is marked as "assume unchanged", there is no
2447          * guarantee that work tree matches what we are looking for.
2448          */
2449         if ((ce->ce_flags & CE_VALID) || ce_skip_worktree(ce))
2450                 return 0;
2452         /*
2453          * If ce matches the file in the work tree, we can reuse it.
2454          */
2455         if (ce_uptodate(ce) ||
2456             (!lstat(name, &st) && !ce_match_stat(ce, &st, 0)))
2457                 return 1;
2459         return 0;
2462 static int populate_from_stdin(struct diff_filespec *s)
2464         struct strbuf buf = STRBUF_INIT;
2465         size_t size = 0;
2467         if (strbuf_read(&buf, 0, 0) < 0)
2468                 return error("error while reading from stdin %s",
2469                                      strerror(errno));
2471         s->should_munmap = 0;
2472         s->data = strbuf_detach(&buf, &size);
2473         s->size = size;
2474         s->should_free = 1;
2475         return 0;
2478 static int diff_populate_gitlink(struct diff_filespec *s, int size_only)
2480         int len;
2481         char *data = xmalloc(100), *dirty = "";
2483         /* Are we looking at the work tree? */
2484         if (s->dirty_submodule)
2485                 dirty = "-dirty";
2487         len = snprintf(data, 100,
2488                        "Subproject commit %s%s\n", sha1_to_hex(s->sha1), dirty);
2489         s->data = data;
2490         s->size = len;
2491         s->should_free = 1;
2492         if (size_only) {
2493                 s->data = NULL;
2494                 free(data);
2495         }
2496         return 0;
2499 /*
2500  * While doing rename detection and pickaxe operation, we may need to
2501  * grab the data for the blob (or file) for our own in-core comparison.
2502  * diff_filespec has data and size fields for this purpose.
2503  */
2504 int diff_populate_filespec(struct diff_filespec *s, int size_only)
2506         int err = 0;
2507         if (!DIFF_FILE_VALID(s))
2508                 die("internal error: asking to populate invalid file.");
2509         if (S_ISDIR(s->mode))
2510                 return -1;
2512         if (s->data)
2513                 return 0;
2515         if (size_only && 0 < s->size)
2516                 return 0;
2518         if (S_ISGITLINK(s->mode))
2519                 return diff_populate_gitlink(s, size_only);
2521         if (!s->sha1_valid ||
2522             reuse_worktree_file(s->path, s->sha1, 0)) {
2523                 struct strbuf buf = STRBUF_INIT;
2524                 struct stat st;
2525                 int fd;
2527                 if (!strcmp(s->path, "-"))
2528                         return populate_from_stdin(s);
2530                 if (lstat(s->path, &st) < 0) {
2531                         if (errno == ENOENT) {
2532                         err_empty:
2533                                 err = -1;
2534                         empty:
2535                                 s->data = (char *)"";
2536                                 s->size = 0;
2537                                 return err;
2538                         }
2539                 }
2540                 s->size = xsize_t(st.st_size);
2541                 if (!s->size)
2542                         goto empty;
2543                 if (S_ISLNK(st.st_mode)) {
2544                         struct strbuf sb = STRBUF_INIT;
2546                         if (strbuf_readlink(&sb, s->path, s->size))
2547                                 goto err_empty;
2548                         s->size = sb.len;
2549                         s->data = strbuf_detach(&sb, NULL);
2550                         s->should_free = 1;
2551                         return 0;
2552                 }
2553                 if (size_only)
2554                         return 0;
2555                 fd = open(s->path, O_RDONLY);
2556                 if (fd < 0)
2557                         goto err_empty;
2558                 s->data = xmmap(NULL, s->size, PROT_READ, MAP_PRIVATE, fd, 0);
2559                 close(fd);
2560                 s->should_munmap = 1;
2562                 /*
2563                  * Convert from working tree format to canonical git format
2564                  */
2565                 if (convert_to_git(s->path, s->data, s->size, &buf, safe_crlf)) {
2566                         size_t size = 0;
2567                         munmap(s->data, s->size);
2568                         s->should_munmap = 0;
2569                         s->data = strbuf_detach(&buf, &size);
2570                         s->size = size;
2571                         s->should_free = 1;
2572                 }
2573         }
2574         else {
2575                 enum object_type type;
2576                 if (size_only) {
2577                         type = sha1_object_info(s->sha1, &s->size);
2578                         if (type < 0)
2579                                 die("unable to read %s", sha1_to_hex(s->sha1));
2580                 } else {
2581                         s->data = read_sha1_file(s->sha1, &type, &s->size);
2582                         if (!s->data)
2583                                 die("unable to read %s", sha1_to_hex(s->sha1));
2584                         s->should_free = 1;
2585                 }
2586         }
2587         return 0;
2590 void diff_free_filespec_blob(struct diff_filespec *s)
2592         if (s->should_free)
2593                 free(s->data);
2594         else if (s->should_munmap)
2595                 munmap(s->data, s->size);
2597         if (s->should_free || s->should_munmap) {
2598                 s->should_free = s->should_munmap = 0;
2599                 s->data = NULL;
2600         }
2603 void diff_free_filespec_data(struct diff_filespec *s)
2605         diff_free_filespec_blob(s);
2606         free(s->cnt_data);
2607         s->cnt_data = NULL;
2610 static void prep_temp_blob(const char *path, struct diff_tempfile *temp,
2611                            void *blob,
2612                            unsigned long size,
2613                            const unsigned char *sha1,
2614                            int mode)
2616         int fd;
2617         struct strbuf buf = STRBUF_INIT;
2618         struct strbuf template = STRBUF_INIT;
2619         char *path_dup = xstrdup(path);
2620         const char *base = basename(path_dup);
2622         /* Generate "XXXXXX_basename.ext" */
2623         strbuf_addstr(&template, "XXXXXX_");
2624         strbuf_addstr(&template, base);
2626         fd = git_mkstemps(temp->tmp_path, PATH_MAX, template.buf,
2627                         strlen(base) + 1);
2628         if (fd < 0)
2629                 die_errno("unable to create temp-file");
2630         if (convert_to_working_tree(path,
2631                         (const char *)blob, (size_t)size, &buf)) {
2632                 blob = buf.buf;
2633                 size = buf.len;
2634         }
2635         if (write_in_full(fd, blob, size) != size)
2636                 die_errno("unable to write temp-file");
2637         close(fd);
2638         temp->name = temp->tmp_path;
2639         strcpy(temp->hex, sha1_to_hex(sha1));
2640         temp->hex[40] = 0;
2641         sprintf(temp->mode, "%06o", mode);
2642         strbuf_release(&buf);
2643         strbuf_release(&template);
2644         free(path_dup);
2647 static struct diff_tempfile *prepare_temp_file(const char *name,
2648                 struct diff_filespec *one)
2650         struct diff_tempfile *temp = claim_diff_tempfile();
2652         if (!DIFF_FILE_VALID(one)) {
2653         not_a_valid_file:
2654                 /* A '-' entry produces this for file-2, and
2655                  * a '+' entry produces this for file-1.
2656                  */
2657                 temp->name = "/dev/null";
2658                 strcpy(temp->hex, ".");
2659                 strcpy(temp->mode, ".");
2660                 return temp;
2661         }
2663         if (!remove_tempfile_installed) {
2664                 atexit(remove_tempfile);
2665                 sigchain_push_common(remove_tempfile_on_signal);
2666                 remove_tempfile_installed = 1;
2667         }
2669         if (!one->sha1_valid ||
2670             reuse_worktree_file(name, one->sha1, 1)) {
2671                 struct stat st;
2672                 if (lstat(name, &st) < 0) {
2673                         if (errno == ENOENT)
2674                                 goto not_a_valid_file;
2675                         die_errno("stat(%s)", name);
2676                 }
2677                 if (S_ISLNK(st.st_mode)) {
2678                         struct strbuf sb = STRBUF_INIT;
2679                         if (strbuf_readlink(&sb, name, st.st_size) < 0)
2680                                 die_errno("readlink(%s)", name);
2681                         prep_temp_blob(name, temp, sb.buf, sb.len,
2682                                        (one->sha1_valid ?
2683                                         one->sha1 : null_sha1),
2684                                        (one->sha1_valid ?
2685                                         one->mode : S_IFLNK));
2686                         strbuf_release(&sb);
2687                 }
2688                 else {
2689                         /* we can borrow from the file in the work tree */
2690                         temp->name = name;
2691                         if (!one->sha1_valid)
2692                                 strcpy(temp->hex, sha1_to_hex(null_sha1));
2693                         else
2694                                 strcpy(temp->hex, sha1_to_hex(one->sha1));
2695                         /* Even though we may sometimes borrow the
2696                          * contents from the work tree, we always want
2697                          * one->mode.  mode is trustworthy even when
2698                          * !(one->sha1_valid), as long as
2699                          * DIFF_FILE_VALID(one).
2700                          */
2701                         sprintf(temp->mode, "%06o", one->mode);
2702                 }
2703                 return temp;
2704         }
2705         else {
2706                 if (diff_populate_filespec(one, 0))
2707                         die("cannot read data blob for %s", one->path);
2708                 prep_temp_blob(name, temp, one->data, one->size,
2709                                one->sha1, one->mode);
2710         }
2711         return temp;
2714 /* An external diff command takes:
2715  *
2716  * diff-cmd name infile1 infile1-sha1 infile1-mode \
2717  *               infile2 infile2-sha1 infile2-mode [ rename-to ]
2718  *
2719  */
2720 static void run_external_diff(const char *pgm,
2721                               const char *name,
2722                               const char *other,
2723                               struct diff_filespec *one,
2724                               struct diff_filespec *two,
2725                               const char *xfrm_msg,
2726                               int complete_rewrite)
2728         const char *spawn_arg[10];
2729         int retval;
2730         const char **arg = &spawn_arg[0];
2732         if (one && two) {
2733                 struct diff_tempfile *temp_one, *temp_two;
2734                 const char *othername = (other ? other : name);
2735                 temp_one = prepare_temp_file(name, one);
2736                 temp_two = prepare_temp_file(othername, two);
2737                 *arg++ = pgm;
2738                 *arg++ = name;
2739                 *arg++ = temp_one->name;
2740                 *arg++ = temp_one->hex;
2741                 *arg++ = temp_one->mode;
2742                 *arg++ = temp_two->name;
2743                 *arg++ = temp_two->hex;
2744                 *arg++ = temp_two->mode;
2745                 if (other) {
2746                         *arg++ = other;
2747                         *arg++ = xfrm_msg;
2748                 }
2749         } else {
2750                 *arg++ = pgm;
2751                 *arg++ = name;
2752         }
2753         *arg = NULL;
2754         fflush(NULL);
2755         retval = run_command_v_opt(spawn_arg, RUN_USING_SHELL);
2756         remove_tempfile();
2757         if (retval) {
2758                 fprintf(stderr, "external diff died, stopping at %s.\n", name);
2759                 exit(1);
2760         }
2763 static int similarity_index(struct diff_filepair *p)
2765         return p->score * 100 / MAX_SCORE;
2768 static void fill_metainfo(struct strbuf *msg,
2769                           const char *name,
2770                           const char *other,
2771                           struct diff_filespec *one,
2772                           struct diff_filespec *two,
2773                           struct diff_options *o,
2774                           struct diff_filepair *p,
2775                           int *must_show_header,
2776                           int use_color)
2778         const char *set = diff_get_color(use_color, DIFF_METAINFO);
2779         const char *reset = diff_get_color(use_color, DIFF_RESET);
2780         struct strbuf *msgbuf;
2781         char *line_prefix = "";
2783         *must_show_header = 1;
2784         if (o->output_prefix) {
2785                 msgbuf = o->output_prefix(o, o->output_prefix_data);
2786                 line_prefix = msgbuf->buf;
2787         }
2788         strbuf_init(msg, PATH_MAX * 2 + 300);
2789         switch (p->status) {
2790         case DIFF_STATUS_COPIED:
2791                 strbuf_addf(msg, "%s%ssimilarity index %d%%",
2792                             line_prefix, set, similarity_index(p));
2793                 strbuf_addf(msg, "%s\n%s%scopy from ",
2794                             reset,  line_prefix, set);
2795                 quote_c_style(name, msg, NULL, 0);
2796                 strbuf_addf(msg, "%s\n%s%scopy to ", reset, line_prefix, set);
2797                 quote_c_style(other, msg, NULL, 0);
2798                 strbuf_addf(msg, "%s\n", reset);
2799                 break;
2800         case DIFF_STATUS_RENAMED:
2801                 strbuf_addf(msg, "%s%ssimilarity index %d%%",
2802                             line_prefix, set, similarity_index(p));
2803                 strbuf_addf(msg, "%s\n%s%srename from ",
2804                             reset, line_prefix, set);
2805                 quote_c_style(name, msg, NULL, 0);
2806                 strbuf_addf(msg, "%s\n%s%srename to ",
2807                             reset, line_prefix, set);
2808                 quote_c_style(other, msg, NULL, 0);
2809                 strbuf_addf(msg, "%s\n", reset);
2810                 break;
2811         case DIFF_STATUS_MODIFIED:
2812                 if (p->score) {
2813                         strbuf_addf(msg, "%s%sdissimilarity index %d%%%s\n",
2814                                     line_prefix,
2815                                     set, similarity_index(p), reset);
2816                         break;
2817                 }
2818                 /* fallthru */
2819         default:
2820                 *must_show_header = 0;
2821         }
2822         if (one && two && hashcmp(one->sha1, two->sha1)) {
2823                 int abbrev = DIFF_OPT_TST(o, FULL_INDEX) ? 40 : DEFAULT_ABBREV;
2825                 if (DIFF_OPT_TST(o, BINARY)) {
2826                         mmfile_t mf;
2827                         if ((!fill_mmfile(&mf, one) && diff_filespec_is_binary(one)) ||
2828                             (!fill_mmfile(&mf, two) && diff_filespec_is_binary(two)))
2829                                 abbrev = 40;
2830                 }
2831                 strbuf_addf(msg, "%s%sindex %s..", line_prefix, set,
2832                             find_unique_abbrev(one->sha1, abbrev));
2833                 strbuf_addstr(msg, find_unique_abbrev(two->sha1, abbrev));
2834                 if (one->mode == two->mode)
2835                         strbuf_addf(msg, " %06o", one->mode);
2836                 strbuf_addf(msg, "%s\n", reset);
2837         }
2840 static void run_diff_cmd(const char *pgm,
2841                          const char *name,
2842                          const char *other,
2843                          const char *attr_path,
2844                          struct diff_filespec *one,
2845                          struct diff_filespec *two,
2846                          struct strbuf *msg,
2847                          struct diff_options *o,
2848                          struct diff_filepair *p)
2850         const char *xfrm_msg = NULL;
2851         int complete_rewrite = (p->status == DIFF_STATUS_MODIFIED) && p->score;
2852         int must_show_header = 0;
2854         if (!DIFF_OPT_TST(o, ALLOW_EXTERNAL))
2855                 pgm = NULL;
2856         else {
2857                 struct userdiff_driver *drv = userdiff_find_by_path(attr_path);
2858                 if (drv && drv->external)
2859                         pgm = drv->external;
2860         }
2862         if (msg) {
2863                 /*
2864                  * don't use colors when the header is intended for an
2865                  * external diff driver
2866                  */
2867                 fill_metainfo(msg, name, other, one, two, o, p,
2868                               &must_show_header,
2869                               want_color(o->use_color) && !pgm);
2870                 xfrm_msg = msg->len ? msg->buf : NULL;
2871         }
2873         if (pgm) {
2874                 run_external_diff(pgm, name, other, one, two, xfrm_msg,
2875                                   complete_rewrite);
2876                 return;
2877         }
2878         if (one && two)
2879                 builtin_diff(name, other ? other : name,
2880                              one, two, xfrm_msg, must_show_header,
2881                              o, complete_rewrite);
2882         else
2883                 fprintf(o->file, "* Unmerged path %s\n", name);
2886 static void diff_fill_sha1_info(struct diff_filespec *one)
2888         if (DIFF_FILE_VALID(one)) {
2889                 if (!one->sha1_valid) {
2890                         struct stat st;
2891                         if (!strcmp(one->path, "-")) {
2892                                 hashcpy(one->sha1, null_sha1);
2893                                 return;
2894                         }
2895                         if (lstat(one->path, &st) < 0)
2896                                 die_errno("stat '%s'", one->path);
2897                         if (index_path(one->sha1, one->path, &st, 0))
2898                                 die("cannot hash %s", one->path);
2899                 }
2900         }
2901         else
2902                 hashclr(one->sha1);
2905 static void strip_prefix(int prefix_length, const char **namep, const char **otherp)
2907         /* Strip the prefix but do not molest /dev/null and absolute paths */
2908         if (*namep && **namep != '/') {
2909                 *namep += prefix_length;
2910                 if (**namep == '/')
2911                         ++*namep;
2912         }
2913         if (*otherp && **otherp != '/') {
2914                 *otherp += prefix_length;
2915                 if (**otherp == '/')
2916                         ++*otherp;
2917         }
2920 static void run_diff(struct diff_filepair *p, struct diff_options *o)
2922         const char *pgm = external_diff();
2923         struct strbuf msg;
2924         struct diff_filespec *one = p->one;
2925         struct diff_filespec *two = p->two;
2926         const char *name;
2927         const char *other;
2928         const char *attr_path;
2930         name  = p->one->path;
2931         other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2932         attr_path = name;
2933         if (o->prefix_length)
2934                 strip_prefix(o->prefix_length, &name, &other);
2936         if (DIFF_PAIR_UNMERGED(p)) {
2937                 run_diff_cmd(pgm, name, NULL, attr_path,
2938                              NULL, NULL, NULL, o, p);
2939                 return;
2940         }
2942         diff_fill_sha1_info(one);
2943         diff_fill_sha1_info(two);
2945         if (!pgm &&
2946             DIFF_FILE_VALID(one) && DIFF_FILE_VALID(two) &&
2947             (S_IFMT & one->mode) != (S_IFMT & two->mode)) {
2948                 /*
2949                  * a filepair that changes between file and symlink
2950                  * needs to be split into deletion and creation.
2951                  */
2952                 struct diff_filespec *null = alloc_filespec(two->path);
2953                 run_diff_cmd(NULL, name, other, attr_path,
2954                              one, null, &msg, o, p);
2955                 free(null);
2956                 strbuf_release(&msg);
2958                 null = alloc_filespec(one->path);
2959                 run_diff_cmd(NULL, name, other, attr_path,
2960                              null, two, &msg, o, p);
2961                 free(null);
2962         }
2963         else
2964                 run_diff_cmd(pgm, name, other, attr_path,
2965                              one, two, &msg, o, p);
2967         strbuf_release(&msg);
2970 static void run_diffstat(struct diff_filepair *p, struct diff_options *o,
2971                          struct diffstat_t *diffstat)
2973         const char *name;
2974         const char *other;
2975         int complete_rewrite = 0;
2977         if (DIFF_PAIR_UNMERGED(p)) {
2978                 /* unmerged */
2979                 builtin_diffstat(p->one->path, NULL, NULL, NULL, diffstat, o, 0);
2980                 return;
2981         }
2983         name = p->one->path;
2984         other = (strcmp(name, p->two->path) ? p->two->path : NULL);
2986         if (o->prefix_length)
2987                 strip_prefix(o->prefix_length, &name, &other);
2989         diff_fill_sha1_info(p->one);
2990         diff_fill_sha1_info(p->two);
2992         if (p->status == DIFF_STATUS_MODIFIED && p->score)
2993                 complete_rewrite = 1;
2994         builtin_diffstat(name, other, p->one, p->two, diffstat, o, complete_rewrite);
2997 static void run_checkdiff(struct diff_filepair *p, struct diff_options *o)
2999         const char *name;
3000         const char *other;
3001         const char *attr_path;
3003         if (DIFF_PAIR_UNMERGED(p)) {
3004                 /* unmerged */
3005                 return;
3006         }
3008         name = p->one->path;
3009         other = (strcmp(name, p->two->path) ? p->two->path : NULL);
3010         attr_path = other ? other : name;
3012         if (o->prefix_length)
3013                 strip_prefix(o->prefix_length, &name, &other);
3015         diff_fill_sha1_info(p->one);
3016         diff_fill_sha1_info(p->two);
3018         builtin_checkdiff(name, other, attr_path, p->one, p->two, o);
3021 void diff_setup(struct diff_options *options)
3023         memcpy(options, &default_diff_options, sizeof(*options));
3025         options->file = stdout;
3027         options->line_termination = '\n';
3028         options->break_opt = -1;
3029         options->rename_limit = -1;
3030         options->dirstat_permille = diff_dirstat_permille_default;
3031         options->context = 3;
3033         options->change = diff_change;
3034         options->add_remove = diff_addremove;
3035         options->use_color = diff_use_color_default;
3036         options->detect_rename = diff_detect_rename_default;
3038         if (diff_no_prefix) {
3039                 options->a_prefix = options->b_prefix = "";
3040         } else if (!diff_mnemonic_prefix) {
3041                 options->a_prefix = "a/";
3042                 options->b_prefix = "b/";
3043         }
3046 int diff_setup_done(struct diff_options *options)
3048         int count = 0;
3050         if (options->output_format & DIFF_FORMAT_NAME)
3051                 count++;
3052         if (options->output_format & DIFF_FORMAT_NAME_STATUS)
3053                 count++;
3054         if (options->output_format & DIFF_FORMAT_CHECKDIFF)
3055                 count++;
3056         if (options->output_format & DIFF_FORMAT_NO_OUTPUT)
3057                 count++;
3058         if (count > 1)
3059                 die("--name-only, --name-status, --check and -s are mutually exclusive");
3061         /*
3062          * Most of the time we can say "there are changes"
3063          * only by checking if there are changed paths, but
3064          * --ignore-whitespace* options force us to look
3065          * inside contents.
3066          */
3068         if (DIFF_XDL_TST(options, IGNORE_WHITESPACE) ||
3069             DIFF_XDL_TST(options, IGNORE_WHITESPACE_CHANGE) ||
3070             DIFF_XDL_TST(options, IGNORE_WHITESPACE_AT_EOL))
3071                 DIFF_OPT_SET(options, DIFF_FROM_CONTENTS);
3072         else
3073                 DIFF_OPT_CLR(options, DIFF_FROM_CONTENTS);
3075         if (DIFF_OPT_TST(options, FIND_COPIES_HARDER))
3076                 options->detect_rename = DIFF_DETECT_COPY;
3078         if (!DIFF_OPT_TST(options, RELATIVE_NAME))
3079                 options->prefix = NULL;
3080         if (options->prefix)
3081                 options->prefix_length = strlen(options->prefix);
3082         else
3083                 options->prefix_length = 0;
3085         if (options->output_format & (DIFF_FORMAT_NAME |
3086                                       DIFF_FORMAT_NAME_STATUS |
3087                                       DIFF_FORMAT_CHECKDIFF |
3088                                       DIFF_FORMAT_NO_OUTPUT))
3089                 options->output_format &= ~(DIFF_FORMAT_RAW |
3090                                             DIFF_FORMAT_NUMSTAT |
3091                                             DIFF_FORMAT_DIFFSTAT |
3092                                             DIFF_FORMAT_SHORTSTAT |
3093                                             DIFF_FORMAT_DIRSTAT |
3094                                             DIFF_FORMAT_SUMMARY |
3095                                             DIFF_FORMAT_PATCH);
3097         /*
3098          * These cases always need recursive; we do not drop caller-supplied
3099          * recursive bits for other formats here.
3100          */
3101         if (options->output_format & (DIFF_FORMAT_PATCH |
3102                                       DIFF_FORMAT_NUMSTAT |
3103                                       DIFF_FORMAT_DIFFSTAT |
3104                                       DIFF_FORMAT_SHORTSTAT |
3105                                       DIFF_FORMAT_DIRSTAT |
3106                                       DIFF_FORMAT_SUMMARY |
3107                                       DIFF_FORMAT_CHECKDIFF))
3108                 DIFF_OPT_SET(options, RECURSIVE);
3109         /*
3110          * Also pickaxe would not work very well if you do not say recursive
3111          */
3112         if (options->pickaxe)
3113                 DIFF_OPT_SET(options, RECURSIVE);
3114         /*
3115          * When patches are generated, submodules diffed against the work tree
3116          * must be checked for dirtiness too so it can be shown in the output
3117          */
3118         if (options->output_format & DIFF_FORMAT_PATCH)
3119                 DIFF_OPT_SET(options, DIRTY_SUBMODULES);
3121         if (options->detect_rename && options->rename_limit < 0)
3122                 options->rename_limit = diff_rename_limit_default;
3123         if (options->setup & DIFF_SETUP_USE_CACHE) {
3124                 if (!active_cache)
3125                         /* read-cache does not die even when it fails
3126                          * so it is safe for us to do this here.  Also
3127                          * it does not smudge active_cache or active_nr
3128                          * when it fails, so we do not have to worry about
3129                          * cleaning it up ourselves either.
3130                          */
3131                         read_cache();
3132         }
3133         if (options->abbrev <= 0 || 40 < options->abbrev)
3134                 options->abbrev = 40; /* full */
3136         /*
3137          * It does not make sense to show the first hit we happened
3138          * to have found.  It does not make sense not to return with
3139          * exit code in such a case either.
3140          */
3141         if (DIFF_OPT_TST(options, QUICK)) {
3142                 options->output_format = DIFF_FORMAT_NO_OUTPUT;
3143                 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3144         }
3146         return 0;
3149 static int opt_arg(const char *arg, int arg_short, const char *arg_long, int *val)
3151         char c, *eq;
3152         int len;
3154         if (*arg != '-')
3155                 return 0;
3156         c = *++arg;
3157         if (!c)
3158                 return 0;
3159         if (c == arg_short) {
3160                 c = *++arg;
3161                 if (!c)
3162                         return 1;
3163                 if (val && isdigit(c)) {
3164                         char *end;
3165                         int n = strtoul(arg, &end, 10);
3166                         if (*end)
3167                                 return 0;
3168                         *val = n;
3169                         return 1;
3170                 }
3171                 return 0;
3172         }
3173         if (c != '-')
3174                 return 0;
3175         arg++;
3176         eq = strchr(arg, '=');
3177         if (eq)
3178                 len = eq - arg;
3179         else
3180                 len = strlen(arg);
3181         if (!len || strncmp(arg, arg_long, len))
3182                 return 0;
3183         if (eq) {
3184                 int n;
3185                 char *end;
3186                 if (!isdigit(*++eq))
3187                         return 0;
3188                 n = strtoul(eq, &end, 10);
3189                 if (*end)
3190                         return 0;
3191                 *val = n;
3192         }
3193         return 1;
3196 static int diff_scoreopt_parse(const char *opt);
3198 static inline int short_opt(char opt, const char **argv,
3199                             const char **optarg)
3201         const char *arg = argv[0];
3202         if (arg[0] != '-' || arg[1] != opt)
3203                 return 0;
3204         if (arg[2] != '\0') {
3205                 *optarg = arg + 2;
3206                 return 1;
3207         }
3208         if (!argv[1])
3209                 die("Option '%c' requires a value", opt);
3210         *optarg = argv[1];
3211         return 2;
3214 int parse_long_opt(const char *opt, const char **argv,
3215                    const char **optarg)
3217         const char *arg = argv[0];
3218         if (arg[0] != '-' || arg[1] != '-')
3219                 return 0;
3220         arg += strlen("--");
3221         if (prefixcmp(arg, opt))
3222                 return 0;
3223         arg += strlen(opt);
3224         if (*arg == '=') { /* sticked form: --option=value */
3225                 *optarg = arg + 1;
3226                 return 1;
3227         }
3228         if (*arg != '\0')
3229                 return 0;
3230         /* separate form: --option value */
3231         if (!argv[1])
3232                 die("Option '--%s' requires a value", opt);
3233         *optarg = argv[1];
3234         return 2;
3237 static int stat_opt(struct diff_options *options, const char **av)
3239         const char *arg = av[0];
3240         char *end;
3241         int width = options->stat_width;
3242         int name_width = options->stat_name_width;
3243         int count = options->stat_count;
3244         int argcount = 1;
3246         arg += strlen("--stat");
3247         end = (char *)arg;
3249         switch (*arg) {
3250         case '-':
3251                 if (!prefixcmp(arg, "-width")) {
3252                         arg += strlen("-width");
3253                         if (*arg == '=')
3254                                 width = strtoul(arg + 1, &end, 10);
3255                         else if (!*arg && !av[1])
3256                                 die("Option '--stat-width' requires a value");
3257                         else if (!*arg) {
3258                                 width = strtoul(av[1], &end, 10);
3259                                 argcount = 2;
3260                         }
3261                 } else if (!prefixcmp(arg, "-name-width")) {
3262                         arg += strlen("-name-width");
3263                         if (*arg == '=')
3264                                 name_width = strtoul(arg + 1, &end, 10);
3265                         else if (!*arg && !av[1])
3266                                 die("Option '--stat-name-width' requires a value");
3267                         else if (!*arg) {
3268                                 name_width = strtoul(av[1], &end, 10);
3269                                 argcount = 2;
3270                         }
3271                 } else if (!prefixcmp(arg, "-count")) {
3272                         arg += strlen("-count");
3273                         if (*arg == '=')
3274                                 count = strtoul(arg + 1, &end, 10);
3275                         else if (!*arg && !av[1])
3276                                 die("Option '--stat-count' requires a value");
3277                         else if (!*arg) {
3278                                 count = strtoul(av[1], &end, 10);
3279                                 argcount = 2;
3280                         }
3281                 }
3282                 break;
3283         case '=':
3284                 width = strtoul(arg+1, &end, 10);
3285                 if (*end == ',')
3286                         name_width = strtoul(end+1, &end, 10);
3287                 if (*end == ',')
3288                         count = strtoul(end+1, &end, 10);
3289         }
3291         /* Important! This checks all the error cases! */
3292         if (*end)
3293                 return 0;
3294         options->output_format |= DIFF_FORMAT_DIFFSTAT;
3295         options->stat_name_width = name_width;
3296         options->stat_width = width;
3297         options->stat_count = count;
3298         return argcount;
3301 static int parse_dirstat_opt(struct diff_options *options, const char *params)
3303         struct strbuf errmsg = STRBUF_INIT;
3304         if (parse_dirstat_params(options, params, &errmsg))
3305                 die(_("Failed to parse --dirstat/-X option parameter:\n%s"),
3306                     errmsg.buf);
3307         strbuf_release(&errmsg);
3308         /*
3309          * The caller knows a dirstat-related option is given from the command
3310          * line; allow it to say "return this_function();"
3311          */
3312         options->output_format |= DIFF_FORMAT_DIRSTAT;
3313         return 1;
3316 int diff_opt_parse(struct diff_options *options, const char **av, int ac)
3318         const char *arg = av[0];
3319         const char *optarg;
3320         int argcount;
3322         /* Output format options */
3323         if (!strcmp(arg, "-p") || !strcmp(arg, "-u") || !strcmp(arg, "--patch"))
3324                 options->output_format |= DIFF_FORMAT_PATCH;
3325         else if (opt_arg(arg, 'U', "unified", &options->context))
3326                 options->output_format |= DIFF_FORMAT_PATCH;
3327         else if (!strcmp(arg, "--raw"))
3328                 options->output_format |= DIFF_FORMAT_RAW;
3329         else if (!strcmp(arg, "--patch-with-raw"))
3330                 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_RAW;
3331         else if (!strcmp(arg, "--numstat"))
3332                 options->output_format |= DIFF_FORMAT_NUMSTAT;
3333         else if (!strcmp(arg, "--shortstat"))
3334                 options->output_format |= DIFF_FORMAT_SHORTSTAT;
3335         else if (!strcmp(arg, "-X") || !strcmp(arg, "--dirstat"))
3336                 return parse_dirstat_opt(options, "");
3337         else if (!prefixcmp(arg, "-X"))
3338                 return parse_dirstat_opt(options, arg + 2);
3339         else if (!prefixcmp(arg, "--dirstat="))
3340                 return parse_dirstat_opt(options, arg + 10);
3341         else if (!strcmp(arg, "--cumulative"))
3342                 return parse_dirstat_opt(options, "cumulative");
3343         else if (!strcmp(arg, "--dirstat-by-file"))
3344                 return parse_dirstat_opt(options, "files");
3345         else if (!prefixcmp(arg, "--dirstat-by-file=")) {
3346                 parse_dirstat_opt(options, "files");
3347                 return parse_dirstat_opt(options, arg + 18);
3348         }
3349         else if (!strcmp(arg, "--check"))
3350                 options->output_format |= DIFF_FORMAT_CHECKDIFF;
3351         else if (!strcmp(arg, "--summary"))
3352                 options->output_format |= DIFF_FORMAT_SUMMARY;
3353         else if (!strcmp(arg, "--patch-with-stat"))
3354                 options->output_format |= DIFF_FORMAT_PATCH | DIFF_FORMAT_DIFFSTAT;
3355         else if (!strcmp(arg, "--name-only"))
3356                 options->output_format |= DIFF_FORMAT_NAME;
3357         else if (!strcmp(arg, "--name-status"))
3358                 options->output_format |= DIFF_FORMAT_NAME_STATUS;
3359         else if (!strcmp(arg, "-s"))
3360                 options->output_format |= DIFF_FORMAT_NO_OUTPUT;
3361         else if (!prefixcmp(arg, "--stat"))
3362                 /* --stat, --stat-width, --stat-name-width, or --stat-count */
3363                 return stat_opt(options, av);
3365         /* renames options */
3366         else if (!prefixcmp(arg, "-B") || !prefixcmp(arg, "--break-rewrites=") ||
3367                  !strcmp(arg, "--break-rewrites")) {
3368                 if ((options->break_opt = diff_scoreopt_parse(arg)) == -1)
3369                         return error("invalid argument to -B: %s", arg+2);
3370         }
3371         else if (!prefixcmp(arg, "-M") || !prefixcmp(arg, "--find-renames=") ||
3372                  !strcmp(arg, "--find-renames")) {
3373                 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3374                         return error("invalid argument to -M: %s", arg+2);
3375                 options->detect_rename = DIFF_DETECT_RENAME;
3376         }
3377         else if (!strcmp(arg, "-D") || !strcmp(arg, "--irreversible-delete")) {
3378                 options->irreversible_delete = 1;
3379         }
3380         else if (!prefixcmp(arg, "-C") || !prefixcmp(arg, "--find-copies=") ||
3381                  !strcmp(arg, "--find-copies")) {
3382                 if (options->detect_rename == DIFF_DETECT_COPY)
3383                         DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3384                 if ((options->rename_score = diff_scoreopt_parse(arg)) == -1)
3385                         return error("invalid argument to -C: %s", arg+2);
3386                 options->detect_rename = DIFF_DETECT_COPY;
3387         }
3388         else if (!strcmp(arg, "--no-renames"))
3389                 options->detect_rename = 0;
3390         else if (!strcmp(arg, "--relative"))
3391                 DIFF_OPT_SET(options, RELATIVE_NAME);
3392         else if (!prefixcmp(arg, "--relative=")) {
3393                 DIFF_OPT_SET(options, RELATIVE_NAME);
3394                 options->prefix = arg + 11;
3395         }
3397         /* xdiff options */
3398         else if (!strcmp(arg, "--minimal"))
3399                 DIFF_XDL_SET(options, NEED_MINIMAL);
3400         else if (!strcmp(arg, "--no-minimal"))
3401                 DIFF_XDL_CLR(options, NEED_MINIMAL);
3402         else if (!strcmp(arg, "-w") || !strcmp(arg, "--ignore-all-space"))
3403                 DIFF_XDL_SET(options, IGNORE_WHITESPACE);
3404         else if (!strcmp(arg, "-b") || !strcmp(arg, "--ignore-space-change"))
3405                 DIFF_XDL_SET(options, IGNORE_WHITESPACE_CHANGE);
3406         else if (!strcmp(arg, "--ignore-space-at-eol"))
3407                 DIFF_XDL_SET(options, IGNORE_WHITESPACE_AT_EOL);
3408         else if (!strcmp(arg, "--patience"))
3409                 DIFF_XDL_SET(options, PATIENCE_DIFF);
3410         else if (!strcmp(arg, "--histogram"))
3411                 DIFF_XDL_SET(options, HISTOGRAM_DIFF);
3413         /* flags options */
3414         else if (!strcmp(arg, "--binary")) {
3415                 options->output_format |= DIFF_FORMAT_PATCH;
3416                 DIFF_OPT_SET(options, BINARY);
3417         }
3418         else if (!strcmp(arg, "--full-index"))
3419                 DIFF_OPT_SET(options, FULL_INDEX);
3420         else if (!strcmp(arg, "-a") || !strcmp(arg, "--text"))
3421                 DIFF_OPT_SET(options, TEXT);
3422         else if (!strcmp(arg, "-R"))
3423                 DIFF_OPT_SET(options, REVERSE_DIFF);
3424         else if (!strcmp(arg, "--find-copies-harder"))
3425                 DIFF_OPT_SET(options, FIND_COPIES_HARDER);
3426         else if (!strcmp(arg, "--follow"))
3427                 DIFF_OPT_SET(options, FOLLOW_RENAMES);
3428         else if (!strcmp(arg, "--color"))
3429                 options->use_color = 1;
3430         else if (!prefixcmp(arg, "--color=")) {
3431                 int value = git_config_colorbool(NULL, arg+8);
3432                 if (value < 0)
3433                         return error("option `color' expects \"always\", \"auto\", or \"never\"");
3434                 options->use_color = value;
3435         }
3436         else if (!strcmp(arg, "--no-color"))
3437                 options->use_color = 0;
3438         else if (!strcmp(arg, "--color-words")) {
3439                 options->use_color = 1;
3440                 options->word_diff = DIFF_WORDS_COLOR;
3441         }
3442         else if (!prefixcmp(arg, "--color-words=")) {
3443                 options->use_color = 1;
3444                 options->word_diff = DIFF_WORDS_COLOR;
3445                 options->word_regex = arg + 14;
3446         }
3447         else if (!strcmp(arg, "--word-diff")) {
3448                 if (options->word_diff == DIFF_WORDS_NONE)
3449                         options->word_diff = DIFF_WORDS_PLAIN;
3450         }
3451         else if (!prefixcmp(arg, "--word-diff=")) {
3452                 const char *type = arg + 12;
3453                 if (!strcmp(type, "plain"))
3454                         options->word_diff = DIFF_WORDS_PLAIN;
3455                 else if (!strcmp(type, "color")) {
3456                         options->use_color = 1;
3457                         options->word_diff = DIFF_WORDS_COLOR;
3458                 }
3459                 else if (!strcmp(type, "porcelain"))
3460                         options->word_diff = DIFF_WORDS_PORCELAIN;
3461                 else if (!strcmp(type, "none"))
3462                         options->word_diff = DIFF_WORDS_NONE;
3463                 else
3464                         die("bad --word-diff argument: %s", type);
3465         }
3466         else if ((argcount = parse_long_opt("word-diff-regex", av, &optarg))) {
3467                 if (options->word_diff == DIFF_WORDS_NONE)
3468                         options->word_diff = DIFF_WORDS_PLAIN;
3469                 options->word_regex = optarg;
3470                 return argcount;
3471         }
3472         else if (!strcmp(arg, "--exit-code"))
3473                 DIFF_OPT_SET(options, EXIT_WITH_STATUS);
3474         else if (!strcmp(arg, "--quiet"))
3475                 DIFF_OPT_SET(options, QUICK);
3476         else if (!strcmp(arg, "--ext-diff"))
3477                 DIFF_OPT_SET(options, ALLOW_EXTERNAL);
3478         else if (!strcmp(arg, "--no-ext-diff"))
3479                 DIFF_OPT_CLR(options, ALLOW_EXTERNAL);
3480         else if (!strcmp(arg, "--textconv"))
3481                 DIFF_OPT_SET(options, ALLOW_TEXTCONV);
3482         else if (!strcmp(arg, "--no-textconv"))
3483                 DIFF_OPT_CLR(options, ALLOW_TEXTCONV);
3484         else if (!strcmp(arg, "--ignore-submodules")) {
3485                 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3486                 handle_ignore_submodules_arg(options, "all");
3487         } else if (!prefixcmp(arg, "--ignore-submodules=")) {
3488                 DIFF_OPT_SET(options, OVERRIDE_SUBMODULE_CONFIG);
3489                 handle_ignore_submodules_arg(options, arg + 20);
3490         } else if (!strcmp(arg, "--submodule"))
3491                 DIFF_OPT_SET(options, SUBMODULE_LOG);
3492         else if (!prefixcmp(arg, "--submodule=")) {
3493                 if (!strcmp(arg + 12, "log"))
3494                         DIFF_OPT_SET(options, SUBMODULE_LOG);
3495         }
3497         /* misc options */
3498         else if (!strcmp(arg, "-z"))
3499                 options->line_termination = 0;
3500         else if ((argcount = short_opt('l', av, &optarg))) {
3501                 options->rename_limit = strtoul(optarg, NULL, 10);
3502                 return argcount;
3503         }
3504         else if ((argcount = short_opt('S', av, &optarg))) {
3505                 options->pickaxe = optarg;
3506                 options->pickaxe_opts |= DIFF_PICKAXE_KIND_S;
3507                 return argcount;
3508         } else if ((argcount = short_opt('G', av, &optarg))) {
3509                 options->pickaxe = optarg;
3510                 options->pickaxe_opts |= DIFF_PICKAXE_KIND_G;
3511                 return argcount;
3512         }
3513         else if (!strcmp(arg, "--pickaxe-all"))
3514                 options->pickaxe_opts |= DIFF_PICKAXE_ALL;
3515         else if (!strcmp(arg, "--pickaxe-regex"))
3516                 options->pickaxe_opts |= DIFF_PICKAXE_REGEX;
3517         else if ((argcount = short_opt('O', av, &optarg))) {
3518                 options->orderfile = optarg;
3519                 return argcount;
3520         }
3521         else if ((argcount = parse_long_opt("diff-filter", av, &optarg))) {
3522                 options->filter = optarg;
3523                 return argcount;
3524         }
3525         else if (!strcmp(arg, "--abbrev"))
3526                 options->abbrev = DEFAULT_ABBREV;
3527         else if (!prefixcmp(arg, "--abbrev=")) {
3528                 options->abbrev = strtoul(arg + 9, NULL, 10);
3529                 if (options->abbrev < MINIMUM_ABBREV)
3530                         options->abbrev = MINIMUM_ABBREV;
3531                 else if (40 < options->abbrev)
3532                         options->abbrev = 40;
3533         }
3534         else if ((argcount = parse_long_opt("src-prefix", av, &optarg))) {
3535                 options->a_prefix = optarg;
3536                 return argcount;
3537         }
3538         else if ((argcount = parse_long_opt("dst-prefix", av, &optarg))) {
3539                 options->b_prefix = optarg;
3540                 return argcount;
3541         }
3542         else if (!strcmp(arg, "--no-prefix"))
3543                 options->a_prefix = options->b_prefix = "";
3544         else if (opt_arg(arg, '\0', "inter-hunk-context",
3545                          &options->interhunkcontext))
3546                 ;
3547         else if (!strcmp(arg, "-W"))
3548                 DIFF_OPT_SET(options, FUNCCONTEXT);
3549         else if (!strcmp(arg, "--function-context"))
3550                 DIFF_OPT_SET(options, FUNCCONTEXT);
3551         else if (!strcmp(arg, "--no-function-context"))
3552                 DIFF_OPT_CLR(options, FUNCCONTEXT);
3553         else if ((argcount = parse_long_opt("output", av, &optarg))) {
3554                 options->file = fopen(optarg, "w");
3555                 if (!options->file)
3556                         die_errno("Could not open '%s'", optarg);
3557                 options->close_file = 1;
3558                 return argcount;
3559         } else
3560                 return 0;
3561         return 1;
3564 int parse_rename_score(const char **cp_p)
3566         unsigned long num, scale;
3567         int ch, dot;
3568         const char *cp = *cp_p;
3570         num = 0;
3571         scale = 1;
3572         dot = 0;
3573         for (;;) {
3574                 ch = *cp;
3575                 if ( !dot && ch == '.' ) {
3576                         scale = 1;
3577                         dot = 1;
3578                 } else if ( ch == '%' ) {
3579                         scale = dot ? scale*100 : 100;
3580                         cp++;   /* % is always at the end */
3581                         break;
3582                 } else if ( ch >= '0' && ch <= '9' ) {
3583                         if ( scale < 100000 ) {
3584                                 scale *= 10;
3585                                 num = (num*10) + (ch-'0');
3586                         }
3587                 } else {
3588                         break;
3589                 }
3590                 cp++;
3591         }
3592         *cp_p = cp;
3594         /* user says num divided by scale and we say internally that
3595          * is MAX_SCORE * num / scale.
3596          */
3597         return (int)((num >= scale) ? MAX_SCORE : (MAX_SCORE * num / scale));
3600 static int diff_scoreopt_parse(const char *opt)
3602         int opt1, opt2, cmd;
3604         if (*opt++ != '-')
3605                 return -1;
3606         cmd = *opt++;
3607         if (cmd == '-') {
3608                 /* convert the long-form arguments into short-form versions */
3609                 if (!prefixcmp(opt, "break-rewrites")) {
3610                         opt += strlen("break-rewrites");
3611                         if (*opt == 0 || *opt++ == '=')
3612                                 cmd = 'B';
3613                 } else if (!prefixcmp(opt, "find-copies")) {
3614                         opt += strlen("find-copies");
3615                         if (*opt == 0 || *opt++ == '=')
3616                                 cmd = 'C';
3617                 } else if (!prefixcmp(opt, "find-renames")) {
3618                         opt += strlen("find-renames");
3619                         if (*opt == 0 || *opt++ == '=')
3620                                 cmd = 'M';
3621                 }
3622         }
3623         if (cmd != 'M' && cmd != 'C' && cmd != 'B')
3624                 return -1; /* that is not a -M, -C nor -B option */
3626         opt1 = parse_rename_score(&opt);
3627         if (cmd != 'B')
3628                 opt2 = 0;
3629         else {
3630                 if (*opt == 0)
3631                         opt2 = 0;
3632                 else if (*opt != '/')
3633                         return -1; /* we expect -B80/99 or -B80 */
3634                 else {
3635                         opt++;
3636                         opt2 = parse_rename_score(&opt);
3637                 }
3638         }
3639         if (*opt != 0)
3640                 return -1;
3641         return opt1 | (opt2 << 16);
3644 struct diff_queue_struct diff_queued_diff;
3646 void diff_q(struct diff_queue_struct *queue, struct diff_filepair *dp)
3648         if (queue->alloc <= queue->nr) {
3649                 queue->alloc = alloc_nr(queue->alloc);
3650                 queue->queue = xrealloc(queue->queue,
3651                                         sizeof(dp) * queue->alloc);
3652         }
3653         queue->queue[queue->nr++] = dp;
3656 struct diff_filepair *diff_queue(struct diff_queue_struct *queue,
3657                                  struct diff_filespec *one,
3658                                  struct diff_filespec *two)
3660         struct diff_filepair *dp = xcalloc(1, sizeof(*dp));
3661         dp->one = one;
3662         dp->two = two;
3663         if (queue)
3664                 diff_q(queue, dp);
3665         return dp;
3668 void diff_free_filepair(struct diff_filepair *p)
3670         free_filespec(p->one);
3671         free_filespec(p->two);
3672         free(p);
3675 /* This is different from find_unique_abbrev() in that
3676  * it stuffs the result with dots for alignment.
3677  */
3678 const char *diff_unique_abbrev(const unsigned char *sha1, int len)
3680         int abblen;
3681         const char *abbrev;
3682         if (len == 40)
3683                 return sha1_to_hex(sha1);
3685         abbrev = find_unique_abbrev(sha1, len);
3686         abblen = strlen(abbrev);
3687         if (abblen < 37) {
3688                 static char hex[41];
3689                 if (len < abblen && abblen <= len + 2)
3690                         sprintf(hex, "%s%.*s", abbrev, len+3-abblen, "..");
3691                 else
3692                         sprintf(hex, "%s...", abbrev);
3693                 return hex;
3694         }
3695         return sha1_to_hex(sha1);
3698 static void diff_flush_raw(struct diff_filepair *p, struct diff_options *opt)
3700         int line_termination = opt->line_termination;
3701         int inter_name_termination = line_termination ? '\t' : '\0';
3702         if (opt->output_prefix) {
3703                 struct strbuf *msg = NULL;
3704                 msg = opt->output_prefix(opt, opt->output_prefix_data);
3705                 fprintf(opt->file, "%s", msg->buf);
3706         }
3708         if (!(opt->output_format & DIFF_FORMAT_NAME_STATUS)) {
3709                 fprintf(opt->file, ":%06o %06o %s ", p->one->mode, p->two->mode,
3710                         diff_unique_abbrev(p->one->sha1, opt->abbrev));
3711                 fprintf(opt->file, "%s ", diff_unique_abbrev(p->two->sha1, opt->abbrev));
3712         }
3713         if (p->score) {
3714                 fprintf(opt->file, "%c%03d%c", p->status, similarity_index(p),
3715                         inter_name_termination);
3716         } else {
3717                 fprintf(opt->file, "%c%c", p->status, inter_name_termination);
3718         }
3720         if (p->status == DIFF_STATUS_COPIED ||
3721             p->status == DIFF_STATUS_RENAMED) {
3722                 const char *name_a, *name_b;
3723                 name_a = p->one->path;
3724                 name_b = p->two->path;
3725                 strip_prefix(opt->prefix_length, &name_a, &name_b);
3726                 write_name_quoted(name_a, opt->file, inter_name_termination);
3727                 write_name_quoted(name_b, opt->file, line_termination);
3728         } else {
3729                 const char *name_a, *name_b;
3730                 name_a = p->one->mode ? p->one->path : p->two->path;
3731                 name_b = NULL;
3732                 strip_prefix(opt->prefix_length, &name_a, &name_b);
3733                 write_name_quoted(name_a, opt->file, line_termination);
3734         }
3737 int diff_unmodified_pair(struct diff_filepair *p)
3739         /* This function is written stricter than necessary to support
3740          * the currently implemented transformers, but the idea is to
3741          * let transformers to produce diff_filepairs any way they want,
3742          * and filter and clean them up here before producing the output.
3743          */
3744         struct diff_filespec *one = p->one, *two = p->two;
3746         if (DIFF_PAIR_UNMERGED(p))
3747                 return 0; /* unmerged is interesting */
3749         /* deletion, addition, mode or type change
3750          * and rename are all interesting.
3751          */
3752         if (DIFF_FILE_VALID(one) != DIFF_FILE_VALID(two) ||
3753             DIFF_PAIR_MODE_CHANGED(p) ||
3754             strcmp(one->path, two->path))
3755                 return 0;
3757         /* both are valid and point at the same path.  that is, we are
3758          * dealing with a change.
3759          */
3760         if (one->sha1_valid && two->sha1_valid &&
3761             !hashcmp(one->sha1, two->sha1) &&
3762             !one->dirty_submodule && !two->dirty_submodule)
3763                 return 1; /* no change */
3764         if (!one->sha1_valid && !two->sha1_valid)
3765                 return 1; /* both look at the same file on the filesystem. */
3766         return 0;
3769 static void diff_flush_patch(struct diff_filepair *p, struct diff_options *o)
3771         if (diff_unmodified_pair(p))
3772                 return;
3774         if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3775             (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3776                 return; /* no tree diffs in patch format */
3778         run_diff(p, o);
3781 static void diff_flush_stat(struct diff_filepair *p, struct diff_options *o,
3782                             struct diffstat_t *diffstat)
3784         if (diff_unmodified_pair(p))
3785                 return;
3787         if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3788             (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3789                 return; /* no useful stat for tree diffs */
3791         run_diffstat(p, o, diffstat);
3794 static void diff_flush_checkdiff(struct diff_filepair *p,
3795                 struct diff_options *o)
3797         if (diff_unmodified_pair(p))
3798                 return;
3800         if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
3801             (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
3802                 return; /* nothing to check in tree diffs */
3804         run_checkdiff(p, o);
3807 int diff_queue_is_empty(void)
3809         struct diff_queue_struct *q = &diff_queued_diff;
3810         int i;
3811         for (i = 0; i < q->nr; i++)
3812                 if (!diff_unmodified_pair(q->queue[i]))
3813                         return 0;
3814         return 1;
3817 #if DIFF_DEBUG
3818 void diff_debug_filespec(struct diff_filespec *s, int x, const char *one)
3820         fprintf(stderr, "queue[%d] %s (%s) %s %06o %s\n",
3821                 x, one ? one : "",
3822                 s->path,
3823                 DIFF_FILE_VALID(s) ? "valid" : "invalid",
3824                 s->mode,
3825                 s->sha1_valid ? sha1_to_hex(s->sha1) : "");
3826         fprintf(stderr, "queue[%d] %s size %lu flags %d\n",
3827                 x, one ? one : "",
3828                 s->size, s->xfrm_flags);
3831 void diff_debug_filepair(const struct diff_filepair *p, int i)
3833         diff_debug_filespec(p->one, i, "one");
3834         diff_debug_filespec(p->two, i, "two");
3835         fprintf(stderr, "score %d, status %c rename_used %d broken %d\n",
3836                 p->score, p->status ? p->status : '?',
3837                 p->one->rename_used, p->broken_pair);
3840 void diff_debug_queue(const char *msg, struct diff_queue_struct *q)
3842         int i;
3843         if (msg)
3844                 fprintf(stderr, "%s\n", msg);
3845         fprintf(stderr, "q->nr = %d\n", q->nr);
3846         for (i = 0; i < q->nr; i++) {
3847                 struct diff_filepair *p = q->queue[i];
3848                 diff_debug_filepair(p, i);
3849         }
3851 #endif
3853 static void diff_resolve_rename_copy(void)
3855         int i;
3856         struct diff_filepair *p;
3857         struct diff_queue_struct *q = &diff_queued_diff;
3859         diff_debug_queue("resolve-rename-copy", q);
3861         for (i = 0; i < q->nr; i++) {
3862                 p = q->queue[i];
3863                 p->status = 0; /* undecided */
3864                 if (DIFF_PAIR_UNMERGED(p))
3865                         p->status = DIFF_STATUS_UNMERGED;
3866                 else if (!DIFF_FILE_VALID(p->one))
3867                         p->status = DIFF_STATUS_ADDED;
3868                 else if (!DIFF_FILE_VALID(p->two))
3869                         p->status = DIFF_STATUS_DELETED;
3870                 else if (DIFF_PAIR_TYPE_CHANGED(p))
3871                         p->status = DIFF_STATUS_TYPE_CHANGED;
3873                 /* from this point on, we are dealing with a pair
3874                  * whose both sides are valid and of the same type, i.e.
3875                  * either in-place edit or rename/copy edit.
3876                  */
3877                 else if (DIFF_PAIR_RENAME(p)) {
3878                         /*
3879                          * A rename might have re-connected a broken
3880                          * pair up, causing the pathnames to be the
3881                          * same again. If so, that's not a rename at
3882                          * all, just a modification..
3883                          *
3884                          * Otherwise, see if this source was used for
3885                          * multiple renames, in which case we decrement
3886                          * the count, and call it a copy.
3887                          */
3888                         if (!strcmp(p->one->path, p->two->path))
3889                                 p->status = DIFF_STATUS_MODIFIED;
3890                         else if (--p->one->rename_used > 0)
3891                                 p->status = DIFF_STATUS_COPIED;
3892                         else
3893                                 p->status = DIFF_STATUS_RENAMED;
3894                 }
3895                 else if (hashcmp(p->one->sha1, p->two->sha1) ||
3896                          p->one->mode != p->two->mode ||
3897                          p->one->dirty_submodule ||
3898                          p->two->dirty_submodule ||
3899                          is_null_sha1(p->one->sha1))
3900                         p->status = DIFF_STATUS_MODIFIED;
3901                 else {
3902                         /* This is a "no-change" entry and should not
3903                          * happen anymore, but prepare for broken callers.
3904                          */
3905                         error("feeding unmodified %s to diffcore",
3906                               p->one->path);
3907                         p->status = DIFF_STATUS_UNKNOWN;
3908                 }
3909         }
3910         diff_debug_queue("resolve-rename-copy done", q);
3913 static int check_pair_status(struct diff_filepair *p)
3915         switch (p->status) {
3916         case DIFF_STATUS_UNKNOWN:
3917                 return 0;
3918         case 0:
3919                 die("internal error in diff-resolve-rename-copy");
3920         default:
3921                 return 1;
3922         }
3925 static void flush_one_pair(struct diff_filepair *p, struct diff_options *opt)
3927         int fmt = opt->output_format;
3929         if (fmt & DIFF_FORMAT_CHECKDIFF)
3930                 diff_flush_checkdiff(p, opt);
3931         else if (fmt & (DIFF_FORMAT_RAW | DIFF_FORMAT_NAME_STATUS))
3932                 diff_flush_raw(p, opt);
3933         else if (fmt & DIFF_FORMAT_NAME) {
3934                 const char *name_a, *name_b;
3935                 name_a = p->two->path;
3936                 name_b = NULL;
3937                 strip_prefix(opt->prefix_length, &name_a, &name_b);
3938                 write_name_quoted(name_a, opt->file, opt->line_termination);
3939         }
3942 static void show_file_mode_name(FILE *file, const char *newdelete, struct diff_filespec *fs)
3944         if (fs->mode)
3945                 fprintf(file, " %s mode %06o ", newdelete, fs->mode);
3946         else
3947                 fprintf(file, " %s ", newdelete);
3948         write_name_quoted(fs->path, file, '\n');
3952 static void show_mode_change(FILE *file, struct diff_filepair *p, int show_name,
3953                 const char *line_prefix)
3955         if (p->one->mode && p->two->mode && p->one->mode != p->two->mode) {
3956                 fprintf(file, "%s mode change %06o => %06o%c", line_prefix, p->one->mode,
3957                         p->two->mode, show_name ? ' ' : '\n');
3958                 if (show_name) {
3959                         write_name_quoted(p->two->path, file, '\n');
3960                 }
3961         }
3964 static void show_rename_copy(FILE *file, const char *renamecopy, struct diff_filepair *p,
3965                         const char *line_prefix)
3967         char *names = pprint_rename(p->one->path, p->two->path);
3969         fprintf(file, " %s %s (%d%%)\n", renamecopy, names, similarity_index(p));
3970         free(names);
3971         show_mode_change(file, p, 0, line_prefix);
3974 static void diff_summary(struct diff_options *opt, struct diff_filepair *p)
3976         FILE *file = opt->file;
3977         char *line_prefix = "";
3979         if (opt->output_prefix) {
3980                 struct strbuf *buf = opt->output_prefix(opt, opt->output_prefix_data);
3981                 line_prefix = buf->buf;
3982         }
3984         switch(p->status) {
3985         case DIFF_STATUS_DELETED:
3986                 fputs(line_prefix, file);
3987                 show_file_mode_name(file, "delete", p->one);
3988                 break;
3989         case DIFF_STATUS_ADDED:
3990                 fputs(line_prefix, file);
3991                 show_file_mode_name(file, "create", p->two);
3992                 break;
3993         case DIFF_STATUS_COPIED:
3994                 fputs(line_prefix, file);
3995                 show_rename_copy(file, "copy", p, line_prefix);
3996                 break;
3997         case DIFF_STATUS_RENAMED:
3998                 fputs(line_prefix, file);
3999                 show_rename_copy(file, "rename", p, line_prefix);
4000                 break;
4001         default:
4002                 if (p->score) {
4003                         fprintf(file, "%s rewrite ", line_prefix);
4004                         write_name_quoted(p->two->path, file, ' ');
4005                         fprintf(file, "(%d%%)\n", similarity_index(p));
4006                 }
4007                 show_mode_change(file, p, !p->score, line_prefix);
4008                 break;
4009         }
4012 struct patch_id_t {
4013         git_SHA_CTX *ctx;
4014         int patchlen;
4015 };
4017 static int remove_space(char *line, int len)
4019         int i;
4020         char *dst = line;
4021         unsigned char c;
4023         for (i = 0; i < len; i++)
4024                 if (!isspace((c = line[i])))
4025                         *dst++ = c;
4027         return dst - line;
4030 static void patch_id_consume(void *priv, char *line, unsigned long len)
4032         struct patch_id_t *data = priv;
4033         int new_len;
4035         /* Ignore line numbers when computing the SHA1 of the patch */
4036         if (!prefixcmp(line, "@@ -"))
4037                 return;
4039         new_len = remove_space(line, len);
4041         git_SHA1_Update(data->ctx, line, new_len);
4042         data->patchlen += new_len;
4045 /* returns 0 upon success, and writes result into sha1 */
4046 static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
4048         struct diff_queue_struct *q = &diff_queued_diff;
4049         int i;
4050         git_SHA_CTX ctx;
4051         struct patch_id_t data;
4052         char buffer[PATH_MAX * 4 + 20];
4054         git_SHA1_Init(&ctx);
4055         memset(&data, 0, sizeof(struct patch_id_t));
4056         data.ctx = &ctx;
4058         for (i = 0; i < q->nr; i++) {
4059                 xpparam_t xpp;
4060                 xdemitconf_t xecfg;
4061                 mmfile_t mf1, mf2;
4062                 struct diff_filepair *p = q->queue[i];
4063                 int len1, len2;
4065                 memset(&xpp, 0, sizeof(xpp));
4066                 memset(&xecfg, 0, sizeof(xecfg));
4067                 if (p->status == 0)
4068                         return error("internal diff status error");
4069                 if (p->status == DIFF_STATUS_UNKNOWN)
4070                         continue;
4071                 if (diff_unmodified_pair(p))
4072                         continue;
4073                 if ((DIFF_FILE_VALID(p->one) && S_ISDIR(p->one->mode)) ||
4074                     (DIFF_FILE_VALID(p->two) && S_ISDIR(p->two->mode)))
4075                         continue;
4076                 if (DIFF_PAIR_UNMERGED(p))
4077                         continue;
4079                 diff_fill_sha1_info(p->one);
4080                 diff_fill_sha1_info(p->two);
4081                 if (fill_mmfile(&mf1, p->one) < 0 ||
4082                                 fill_mmfile(&mf2, p->two) < 0)
4083                         return error("unable to read files to diff");
4085                 len1 = remove_space(p->one->path, strlen(p->one->path));
4086                 len2 = remove_space(p->two->path, strlen(p->two->path));
4087                 if (p->one->mode == 0)
4088                         len1 = snprintf(buffer, sizeof(buffer),
4089                                         "diff--gita/%.*sb/%.*s"
4090                                         "newfilemode%06o"
4091                                         "---/dev/null"
4092                                         "+++b/%.*s",
4093                                         len1, p->one->path,
4094                                         len2, p->two->path,
4095                                         p->two->mode,
4096                                         len2, p->two->path);
4097                 else if (p->two->mode == 0)
4098                         len1 = snprintf(buffer, sizeof(buffer),
4099                                         "diff--gita/%.*sb/%.*s"
4100                                         "deletedfilemode%06o"
4101                                         "---a/%.*s"
4102                                         "+++/dev/null",
4103                                         len1, p->one->path,
4104                                         len2, p->two->path,
4105                                         p->one->mode,
4106                                         len1, p->one->path);
4107                 else
4108                         len1 = snprintf(buffer, sizeof(buffer),
4109                                         "diff--gita/%.*sb/%.*s"
4110                                         "---a/%.*s"
4111                                         "+++b/%.*s",
4112                                         len1, p->one->path,
4113                                         len2, p->two->path,
4114                                         len1, p->one->path,
4115                                         len2, p->two->path);
4116                 git_SHA1_Update(&ctx, buffer, len1);
4118                 if (diff_filespec_is_binary(p->one) ||
4119                     diff_filespec_is_binary(p->two)) {
4120                         git_SHA1_Update(&ctx, sha1_to_hex(p->one->sha1), 40);
4121                         git_SHA1_Update(&ctx, sha1_to_hex(p->two->sha1), 40);
4122                         continue;
4123                 }
4125                 xpp.flags = 0;
4126                 xecfg.ctxlen = 3;
4127                 xecfg.flags = 0;
4128                 xdi_diff_outf(&mf1, &mf2, patch_id_consume, &data,
4129                               &xpp, &xecfg);
4130         }
4132         git_SHA1_Final(sha1, &ctx);
4133         return 0;
4136 int diff_flush_patch_id(struct diff_options *options, unsigned char *sha1)
4138         struct diff_queue_struct *q = &diff_queued_diff;
4139         int i;
4140         int result = diff_get_patch_id(options, sha1);
4142         for (i = 0; i < q->nr; i++)
4143                 diff_free_filepair(q->queue[i]);
4145         free(q->queue);
4146         DIFF_QUEUE_CLEAR(q);
4148         return result;
4151 static int is_summary_empty(const struct diff_queue_struct *q)
4153         int i;
4155         for (i = 0; i < q->nr; i++) {
4156                 const struct diff_filepair *p = q->queue[i];
4158                 switch (p->status) {
4159                 case DIFF_STATUS_DELETED:
4160                 case DIFF_STATUS_ADDED:
4161                 case DIFF_STATUS_COPIED:
4162                 case DIFF_STATUS_RENAMED:
4163                         return 0;
4164                 default:
4165                         if (p->score)
4166                                 return 0;
4167                         if (p->one->mode && p->two->mode &&
4168                             p->one->mode != p->two->mode)
4169                                 return 0;
4170                         break;
4171                 }
4172         }
4173         return 1;
4176 static const char rename_limit_warning[] =
4177 "inexact rename detection was skipped due to too many files.";
4179 static const char degrade_cc_to_c_warning[] =
4180 "only found copies from modified paths due to too many files.";
4182 static const char rename_limit_advice[] =
4183 "you may want to set your %s variable to at least "
4184 "%d and retry the command.";
4186 void diff_warn_rename_limit(const char *varname, int needed, int degraded_cc)
4188         if (degraded_cc)
4189                 warning(degrade_cc_to_c_warning);
4190         else if (needed)
4191                 warning(rename_limit_warning);
4192         else
4193                 return;
4194         if (0 < needed && needed < 32767)
4195                 warning(rename_limit_advice, varname, needed);
4198 void diff_flush(struct diff_options *options)
4200         struct diff_queue_struct *q = &diff_queued_diff;
4201         int i, output_format = options->output_format;
4202         int separator = 0;
4203         int dirstat_by_line = 0;
4205         /*
4206          * Order: raw, stat, summary, patch
4207          * or:    name/name-status/checkdiff (other bits clear)
4208          */
4209         if (!q->nr)
4210                 goto free_queue;
4212         if (output_format & (DIFF_FORMAT_RAW |
4213                              DIFF_FORMAT_NAME |
4214                              DIFF_FORMAT_NAME_STATUS |
4215                              DIFF_FORMAT_CHECKDIFF)) {
4216                 for (i = 0; i < q->nr; i++) {
4217                         struct diff_filepair *p = q->queue[i];
4218                         if (check_pair_status(p))
4219                                 flush_one_pair(p, options);
4220                 }
4221                 separator++;
4222         }
4224         if (output_format & DIFF_FORMAT_DIRSTAT && DIFF_OPT_TST(options, DIRSTAT_BY_LINE))
4225                 dirstat_by_line = 1;
4227         if (output_format & (DIFF_FORMAT_DIFFSTAT|DIFF_FORMAT_SHORTSTAT|DIFF_FORMAT_NUMSTAT) ||
4228             dirstat_by_line) {
4229                 struct diffstat_t diffstat;
4231                 memset(&diffstat, 0, sizeof(struct diffstat_t));
4232                 for (i = 0; i < q->nr; i++) {
4233                         struct diff_filepair *p = q->queue[i];
4234                         if (check_pair_status(p))
4235                                 diff_flush_stat(p, options, &diffstat);
4236                 }
4237                 if (output_format & DIFF_FORMAT_NUMSTAT)
4238                         show_numstat(&diffstat, options);
4239                 if (output_format & DIFF_FORMAT_DIFFSTAT)
4240                         show_stats(&diffstat, options);
4241                 if (output_format & DIFF_FORMAT_SHORTSTAT)
4242                         show_shortstats(&diffstat, options);
4243                 if (output_format & DIFF_FORMAT_DIRSTAT)
4244                         show_dirstat_by_line(&diffstat, options);
4245                 free_diffstat_info(&diffstat);
4246                 separator++;
4247         }
4248         if ((output_format & DIFF_FORMAT_DIRSTAT) && !dirstat_by_line)
4249                 show_dirstat(options);
4251         if (output_format & DIFF_FORMAT_SUMMARY && !is_summary_empty(q)) {
4252                 for (i = 0; i < q->nr; i++) {
4253                         diff_summary(options, q->queue[i]);
4254                 }
4255                 separator++;
4256         }
4258         if (output_format & DIFF_FORMAT_NO_OUTPUT &&
4259             DIFF_OPT_TST(options, EXIT_WITH_STATUS) &&
4260             DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4261                 /*
4262                  * run diff_flush_patch for the exit status. setting
4263                  * options->file to /dev/null should be safe, becaue we
4264                  * aren't supposed to produce any output anyway.
4265                  */
4266                 if (options->close_file)
4267                         fclose(options->file);
4268                 options->file = fopen("/dev/null", "w");
4269                 if (!options->file)
4270                         die_errno("Could not open /dev/null");
4271                 options->close_file = 1;
4272                 for (i = 0; i < q->nr; i++) {
4273                         struct diff_filepair *p = q->queue[i];
4274                         if (check_pair_status(p))
4275                                 diff_flush_patch(p, options);
4276                         if (options->found_changes)
4277                                 break;
4278                 }
4279         }
4281         if (output_format & DIFF_FORMAT_PATCH) {
4282                 if (separator) {
4283                         putc(options->line_termination, options->file);
4284                         if (options->stat_sep) {
4285                                 /* attach patch instead of inline */
4286                                 fputs(options->stat_sep, options->file);
4287                         }
4288                 }
4290                 for (i = 0; i < q->nr; i++) {
4291                         struct diff_filepair *p = q->queue[i];
4292                         if (check_pair_status(p))
4293                                 diff_flush_patch(p, options);
4294                 }
4295         }
4297         if (output_format & DIFF_FORMAT_CALLBACK)
4298                 options->format_callback(q, options, options->format_callback_data);
4300         for (i = 0; i < q->nr; i++)
4301                 diff_free_filepair(q->queue[i]);
4302 free_queue:
4303         free(q->queue);
4304         DIFF_QUEUE_CLEAR(q);
4305         if (options->close_file)
4306                 fclose(options->file);
4308         /*
4309          * Report the content-level differences with HAS_CHANGES;
4310          * diff_addremove/diff_change does not set the bit when
4311          * DIFF_FROM_CONTENTS is in effect (e.g. with -w).
4312          */
4313         if (DIFF_OPT_TST(options, DIFF_FROM_CONTENTS)) {
4314                 if (options->found_changes)
4315                         DIFF_OPT_SET(options, HAS_CHANGES);
4316                 else
4317                         DIFF_OPT_CLR(options, HAS_CHANGES);
4318         }
4321 static void diffcore_apply_filter(const char *filter)
4323         int i;
4324         struct diff_queue_struct *q = &diff_queued_diff;
4325         struct diff_queue_struct outq;
4326         DIFF_QUEUE_CLEAR(&outq);
4328         if (!filter)
4329                 return;
4331         if (strchr(filter, DIFF_STATUS_FILTER_AON)) {
4332                 int found;
4333                 for (i = found = 0; !found && i < q->nr; i++) {
4334                         struct diff_filepair *p = q->queue[i];
4335                         if (((p->status == DIFF_STATUS_MODIFIED) &&
4336                              ((p->score &&
4337                                strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4338                               (!p->score &&
4339                                strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4340                             ((p->status != DIFF_STATUS_MODIFIED) &&
4341                              strchr(filter, p->status)))
4342                                 found++;
4343                 }
4344                 if (found)
4345                         return;
4347                 /* otherwise we will clear the whole queue
4348                  * by copying the empty outq at the end of this
4349                  * function, but first clear the current entries
4350                  * in the queue.
4351                  */
4352                 for (i = 0; i < q->nr; i++)
4353                         diff_free_filepair(q->queue[i]);
4354         }
4355         else {
4356                 /* Only the matching ones */
4357                 for (i = 0; i < q->nr; i++) {
4358                         struct diff_filepair *p = q->queue[i];
4360                         if (((p->status == DIFF_STATUS_MODIFIED) &&
4361                              ((p->score &&
4362                                strchr(filter, DIFF_STATUS_FILTER_BROKEN)) ||
4363                               (!p->score &&
4364                                strchr(filter, DIFF_STATUS_MODIFIED)))) ||
4365                             ((p->status != DIFF_STATUS_MODIFIED) &&
4366                              strchr(filter, p->status)))
4367                                 diff_q(&outq, p);
4368                         else
4369                                 diff_free_filepair(p);
4370                 }
4371         }
4372         free(q->queue);
4373         *q = outq;
4376 /* Check whether two filespecs with the same mode and size are identical */
4377 static int diff_filespec_is_identical(struct diff_filespec *one,
4378                                       struct diff_filespec *two)
4380         if (S_ISGITLINK(one->mode))
4381                 return 0;
4382         if (diff_populate_filespec(one, 0))
4383                 return 0;
4384         if (diff_populate_filespec(two, 0))
4385                 return 0;
4386         return !memcmp(one->data, two->data, one->size);
4389 static void diffcore_skip_stat_unmatch(struct diff_options *diffopt)
4391         int i;
4392         struct diff_queue_struct *q = &diff_queued_diff;
4393         struct diff_queue_struct outq;
4394         DIFF_QUEUE_CLEAR(&outq);
4396         for (i = 0; i < q->nr; i++) {
4397                 struct diff_filepair *p = q->queue[i];
4399                 /*
4400                  * 1. Entries that come from stat info dirtiness
4401                  *    always have both sides (iow, not create/delete),
4402                  *    one side of the object name is unknown, with
4403                  *    the same mode and size.  Keep the ones that
4404                  *    do not match these criteria.  They have real
4405                  *    differences.
4406                  *
4407                  * 2. At this point, the file is known to be modified,
4408                  *    with the same mode and size, and the object
4409                  *    name of one side is unknown.  Need to inspect
4410                  *    the identical contents.
4411                  */
4412                 if (!DIFF_FILE_VALID(p->one) || /* (1) */
4413                     !DIFF_FILE_VALID(p->two) ||
4414                     (p->one->sha1_valid && p->two->sha1_valid) ||
4415                     (p->one->mode != p->two->mode) ||
4416                     diff_populate_filespec(p->one, 1) ||
4417                     diff_populate_filespec(p->two, 1) ||
4418                     (p->one->size != p->two->size) ||
4419                     !diff_filespec_is_identical(p->one, p->two)) /* (2) */
4420                         diff_q(&outq, p);
4421                 else {
4422                         /*
4423                          * The caller can subtract 1 from skip_stat_unmatch
4424                          * to determine how many paths were dirty only
4425                          * due to stat info mismatch.
4426                          */
4427                         if (!DIFF_OPT_TST(diffopt, NO_INDEX))
4428                                 diffopt->skip_stat_unmatch++;
4429                         diff_free_filepair(p);
4430                 }
4431         }
4432         free(q->queue);
4433         *q = outq;
4436 static int diffnamecmp(const void *a_, const void *b_)
4438         const struct diff_filepair *a = *((const struct diff_filepair **)a_);
4439         const struct diff_filepair *b = *((const struct diff_filepair **)b_);
4440         const char *name_a, *name_b;
4442         name_a = a->one ? a->one->path : a->two->path;
4443         name_b = b->one ? b->one->path : b->two->path;
4444         return strcmp(name_a, name_b);
4447 void diffcore_fix_diff_index(struct diff_options *options)
4449         struct diff_queue_struct *q = &diff_queued_diff;
4450         qsort(q->queue, q->nr, sizeof(q->queue[0]), diffnamecmp);
4453 void diffcore_std(struct diff_options *options)
4455         if (options->skip_stat_unmatch)
4456                 diffcore_skip_stat_unmatch(options);
4457         if (!options->found_follow) {
4458                 /* See try_to_follow_renames() in tree-diff.c */
4459                 if (options->break_opt != -1)
4460                         diffcore_break(options->break_opt);
4461                 if (options->detect_rename)
4462                         diffcore_rename(options);
4463                 if (options->break_opt != -1)
4464                         diffcore_merge_broken();
4465         }
4466         if (options->pickaxe)
4467                 diffcore_pickaxe(options);
4468         if (options->orderfile)
4469                 diffcore_order(options->orderfile);
4470         if (!options->found_follow)
4471                 /* See try_to_follow_renames() in tree-diff.c */
4472                 diff_resolve_rename_copy();
4473         diffcore_apply_filter(options->filter);
4475         if (diff_queued_diff.nr && !DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4476                 DIFF_OPT_SET(options, HAS_CHANGES);
4477         else
4478                 DIFF_OPT_CLR(options, HAS_CHANGES);
4480         options->found_follow = 0;
4483 int diff_result_code(struct diff_options *opt, int status)
4485         int result = 0;
4487         diff_warn_rename_limit("diff.renamelimit",
4488                                opt->needed_rename_limit,
4489                                opt->degraded_cc_to_c);
4490         if (!DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4491             !(opt->output_format & DIFF_FORMAT_CHECKDIFF))
4492                 return status;
4493         if (DIFF_OPT_TST(opt, EXIT_WITH_STATUS) &&
4494             DIFF_OPT_TST(opt, HAS_CHANGES))
4495                 result |= 01;
4496         if ((opt->output_format & DIFF_FORMAT_CHECKDIFF) &&
4497             DIFF_OPT_TST(opt, CHECK_FAILED))
4498                 result |= 02;
4499         return result;
4502 int diff_can_quit_early(struct diff_options *opt)
4504         return (DIFF_OPT_TST(opt, QUICK) &&
4505                 !opt->filter &&
4506                 DIFF_OPT_TST(opt, HAS_CHANGES));
4509 /*
4510  * Shall changes to this submodule be ignored?
4511  *
4512  * Submodule changes can be configured to be ignored separately for each path,
4513  * but that configuration can be overridden from the command line.
4514  */
4515 static int is_submodule_ignored(const char *path, struct diff_options *options)
4517         int ignored = 0;
4518         unsigned orig_flags = options->flags;
4519         if (!DIFF_OPT_TST(options, OVERRIDE_SUBMODULE_CONFIG))
4520                 set_diffopt_flags_from_submodule_config(options, path);
4521         if (DIFF_OPT_TST(options, IGNORE_SUBMODULES))
4522                 ignored = 1;
4523         options->flags = orig_flags;
4524         return ignored;
4527 void diff_addremove(struct diff_options *options,
4528                     int addremove, unsigned mode,
4529                     const unsigned char *sha1,
4530                     const char *concatpath, unsigned dirty_submodule)
4532         struct diff_filespec *one, *two;
4534         if (S_ISGITLINK(mode) && is_submodule_ignored(concatpath, options))
4535                 return;
4537         /* This may look odd, but it is a preparation for
4538          * feeding "there are unchanged files which should
4539          * not produce diffs, but when you are doing copy
4540          * detection you would need them, so here they are"
4541          * entries to the diff-core.  They will be prefixed
4542          * with something like '=' or '*' (I haven't decided
4543          * which but should not make any difference).
4544          * Feeding the same new and old to diff_change()
4545          * also has the same effect.
4546          * Before the final output happens, they are pruned after
4547          * merged into rename/copy pairs as appropriate.
4548          */
4549         if (DIFF_OPT_TST(options, REVERSE_DIFF))
4550                 addremove = (addremove == '+' ? '-' :
4551                              addremove == '-' ? '+' : addremove);
4553         if (options->prefix &&
4554             strncmp(concatpath, options->prefix, options->prefix_length))
4555                 return;
4557         one = alloc_filespec(concatpath);
4558         two = alloc_filespec(concatpath);
4560         if (addremove != '+')
4561                 fill_filespec(one, sha1, mode);
4562         if (addremove != '-') {
4563                 fill_filespec(two, sha1, mode);
4564                 two->dirty_submodule = dirty_submodule;
4565         }
4567         diff_queue(&diff_queued_diff, one, two);
4568         if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4569                 DIFF_OPT_SET(options, HAS_CHANGES);
4572 void diff_change(struct diff_options *options,
4573                  unsigned old_mode, unsigned new_mode,
4574                  const unsigned char *old_sha1,
4575                  const unsigned char *new_sha1,
4576                  const char *concatpath,
4577                  unsigned old_dirty_submodule, unsigned new_dirty_submodule)
4579         struct diff_filespec *one, *two;
4581         if (S_ISGITLINK(old_mode) && S_ISGITLINK(new_mode) &&
4582             is_submodule_ignored(concatpath, options))
4583                 return;
4585         if (DIFF_OPT_TST(options, REVERSE_DIFF)) {
4586                 unsigned tmp;
4587                 const unsigned char *tmp_c;
4588                 tmp = old_mode; old_mode = new_mode; new_mode = tmp;
4589                 tmp_c = old_sha1; old_sha1 = new_sha1; new_sha1 = tmp_c;
4590                 tmp = old_dirty_submodule; old_dirty_submodule = new_dirty_submodule;
4591                         new_dirty_submodule = tmp;
4592         }
4594         if (options->prefix &&
4595             strncmp(concatpath, options->prefix, options->prefix_length))
4596                 return;
4598         one = alloc_filespec(concatpath);
4599         two = alloc_filespec(concatpath);
4600         fill_filespec(one, old_sha1, old_mode);
4601         fill_filespec(two, new_sha1, new_mode);
4602         one->dirty_submodule = old_dirty_submodule;
4603         two->dirty_submodule = new_dirty_submodule;
4605         diff_queue(&diff_queued_diff, one, two);
4606         if (!DIFF_OPT_TST(options, DIFF_FROM_CONTENTS))
4607                 DIFF_OPT_SET(options, HAS_CHANGES);
4610 struct diff_filepair *diff_unmerge(struct diff_options *options, const char *path)
4612         struct diff_filepair *pair;
4613         struct diff_filespec *one, *two;
4615         if (options->prefix &&
4616             strncmp(path, options->prefix, options->prefix_length))
4617                 return NULL;
4619         one = alloc_filespec(path);
4620         two = alloc_filespec(path);
4621         pair = diff_queue(&diff_queued_diff, one, two);
4622         pair->is_unmerged = 1;
4623         return pair;
4626 static char *run_textconv(const char *pgm, struct diff_filespec *spec,
4627                 size_t *outsize)
4629         struct diff_tempfile *temp;
4630         const char *argv[3];
4631         const char **arg = argv;
4632         struct child_process child;
4633         struct strbuf buf = STRBUF_INIT;
4634         int err = 0;
4636         temp = prepare_temp_file(spec->path, spec);
4637         *arg++ = pgm;
4638         *arg++ = temp->name;
4639         *arg = NULL;
4641         memset(&child, 0, sizeof(child));
4642         child.use_shell = 1;
4643         child.argv = argv;
4644         child.out = -1;
4645         if (start_command(&child)) {
4646                 remove_tempfile();
4647                 return NULL;
4648         }
4650         if (strbuf_read(&buf, child.out, 0) < 0)
4651                 err = error("error reading from textconv command '%s'", pgm);
4652         close(child.out);
4654         if (finish_command(&child) || err) {
4655                 strbuf_release(&buf);
4656                 remove_tempfile();
4657                 return NULL;
4658         }
4659         remove_tempfile();
4661         return strbuf_detach(&buf, outsize);
4664 size_t fill_textconv(struct userdiff_driver *driver,
4665                      struct diff_filespec *df,
4666                      char **outbuf)
4668         size_t size;
4670         if (!driver || !driver->textconv) {
4671                 if (!DIFF_FILE_VALID(df)) {
4672                         *outbuf = "";
4673                         return 0;
4674                 }
4675                 if (diff_populate_filespec(df, 0))
4676                         die("unable to read files to diff");
4677                 *outbuf = df->data;
4678                 return df->size;
4679         }
4681         if (driver->textconv_cache && df->sha1_valid) {
4682                 *outbuf = notes_cache_get(driver->textconv_cache, df->sha1,
4683                                           &size);
4684                 if (*outbuf)
4685                         return size;
4686         }
4688         *outbuf = run_textconv(driver->textconv, df, &size);
4689         if (!*outbuf)
4690                 die("unable to read files to diff");
4692         if (driver->textconv_cache && df->sha1_valid) {
4693                 /* ignore errors, as we might be in a readonly repository */
4694                 notes_cache_put(driver->textconv_cache, df->sha1, *outbuf,
4695                                 size);
4696                 /*
4697                  * we could save up changes and flush them all at the end,
4698                  * but we would need an extra call after all diffing is done.
4699                  * Since generating a cache entry is the slow path anyway,
4700                  * this extra overhead probably isn't a big deal.
4701                  */
4702                 notes_cache_write(driver->textconv_cache);
4703         }
4705         return size;