]> git.ipfire.org Git - thirdparty/git.git/blob - convert.c
convert.c: refactor crlf_action
[thirdparty/git.git] / convert.c
1 #include "cache.h"
2 #include "attr.h"
3 #include "run-command.h"
4 #include "quote.h"
5 #include "sigchain.h"
6
7 /*
8 * convert.c - convert a file when checking it out and checking it in.
9 *
10 * This should use the pathname to decide on whether it wants to do some
11 * more interesting conversions (automatic gzip/unzip, general format
12 * conversions etc etc), but by default it just does automatic CRLF<->LF
13 * translation when the "text" attribute or "auto_crlf" option is set.
14 */
15
16 /* Stat bits: When BIN is set, the txt bits are unset */
17 #define CONVERT_STAT_BITS_TXT_LF 0x1
18 #define CONVERT_STAT_BITS_TXT_CRLF 0x2
19 #define CONVERT_STAT_BITS_BIN 0x4
20
21 enum crlf_action {
22 CRLF_UNDEFINED,
23 CRLF_BINARY,
24 CRLF_TEXT,
25 CRLF_TEXT_INPUT,
26 CRLF_TEXT_CRLF,
27 CRLF_AUTO,
28 CRLF_AUTO_INPUT,
29 CRLF_AUTO_CRLF
30 };
31
32 struct text_stat {
33 /* NUL, CR, LF and CRLF counts */
34 unsigned nul, cr, lf, crlf;
35
36 /* These are just approximations! */
37 unsigned printable, nonprintable;
38 };
39
40 static void gather_stats(const char *buf, unsigned long size, struct text_stat *stats)
41 {
42 unsigned long i;
43
44 memset(stats, 0, sizeof(*stats));
45
46 for (i = 0; i < size; i++) {
47 unsigned char c = buf[i];
48 if (c == '\r') {
49 stats->cr++;
50 if (i+1 < size && buf[i+1] == '\n')
51 stats->crlf++;
52 continue;
53 }
54 if (c == '\n') {
55 stats->lf++;
56 continue;
57 }
58 if (c == 127)
59 /* DEL */
60 stats->nonprintable++;
61 else if (c < 32) {
62 switch (c) {
63 /* BS, HT, ESC and FF */
64 case '\b': case '\t': case '\033': case '\014':
65 stats->printable++;
66 break;
67 case 0:
68 stats->nul++;
69 /* fall through */
70 default:
71 stats->nonprintable++;
72 }
73 }
74 else
75 stats->printable++;
76 }
77
78 /* If file ends with EOF then don't count this EOF as non-printable. */
79 if (size >= 1 && buf[size-1] == '\032')
80 stats->nonprintable--;
81 }
82
83 /*
84 * The same heuristics as diff.c::mmfile_is_binary()
85 * We treat files with bare CR as binary
86 */
87 static int convert_is_binary(unsigned long size, const struct text_stat *stats)
88 {
89 if (stats->cr != stats->crlf)
90 return 1;
91 if (stats->nul)
92 return 1;
93 if ((stats->printable >> 7) < stats->nonprintable)
94 return 1;
95 return 0;
96 }
97
98 static unsigned int gather_convert_stats(const char *data, unsigned long size)
99 {
100 struct text_stat stats;
101 if (!data || !size)
102 return 0;
103 gather_stats(data, size, &stats);
104 if (convert_is_binary(size, &stats))
105 return CONVERT_STAT_BITS_BIN;
106 else if (stats.crlf && stats.crlf == stats.lf)
107 return CONVERT_STAT_BITS_TXT_CRLF;
108 else if (stats.crlf && stats.lf)
109 return CONVERT_STAT_BITS_TXT_CRLF | CONVERT_STAT_BITS_TXT_LF;
110 else if (stats.lf)
111 return CONVERT_STAT_BITS_TXT_LF;
112 else
113 return 0;
114 }
115
116 static const char *gather_convert_stats_ascii(const char *data, unsigned long size)
117 {
118 unsigned int convert_stats = gather_convert_stats(data, size);
119
120 if (convert_stats & CONVERT_STAT_BITS_BIN)
121 return "-text";
122 switch (convert_stats) {
123 case CONVERT_STAT_BITS_TXT_LF:
124 return "lf";
125 case CONVERT_STAT_BITS_TXT_CRLF:
126 return "crlf";
127 case CONVERT_STAT_BITS_TXT_LF | CONVERT_STAT_BITS_TXT_CRLF:
128 return "mixed";
129 default:
130 return "none";
131 }
132 }
133
134 const char *get_cached_convert_stats_ascii(const char *path)
135 {
136 const char *ret;
137 unsigned long sz;
138 void *data = read_blob_data_from_cache(path, &sz);
139 ret = gather_convert_stats_ascii(data, sz);
140 free(data);
141 return ret;
142 }
143
144 const char *get_wt_convert_stats_ascii(const char *path)
145 {
146 const char *ret = "";
147 struct strbuf sb = STRBUF_INIT;
148 if (strbuf_read_file(&sb, path, 0) >= 0)
149 ret = gather_convert_stats_ascii(sb.buf, sb.len);
150 strbuf_release(&sb);
151 return ret;
152 }
153
154 static int text_eol_is_crlf(void)
155 {
156 if (auto_crlf == AUTO_CRLF_TRUE)
157 return 1;
158 else if (auto_crlf == AUTO_CRLF_INPUT)
159 return 0;
160 if (core_eol == EOL_CRLF)
161 return 1;
162 if (core_eol == EOL_UNSET && EOL_NATIVE == EOL_CRLF)
163 return 1;
164 return 0;
165 }
166
167 static enum eol output_eol(enum crlf_action crlf_action)
168 {
169 switch (crlf_action) {
170 case CRLF_BINARY:
171 return EOL_UNSET;
172 case CRLF_TEXT_CRLF:
173 return EOL_CRLF;
174 case CRLF_TEXT_INPUT:
175 return EOL_LF;
176 case CRLF_UNDEFINED:
177 case CRLF_AUTO_CRLF:
178 case CRLF_AUTO_INPUT:
179 case CRLF_TEXT:
180 case CRLF_AUTO:
181 /* fall through */
182 return text_eol_is_crlf() ? EOL_CRLF : EOL_LF;
183 }
184 warning("Illegal crlf_action %d\n", (int)crlf_action);
185 return core_eol;
186 }
187
188 static void check_safe_crlf(const char *path, enum crlf_action crlf_action,
189 struct text_stat *stats, enum safe_crlf checksafe)
190 {
191 if (!checksafe)
192 return;
193
194 if (output_eol(crlf_action) == EOL_LF) {
195 /*
196 * CRLFs would not be restored by checkout:
197 * check if we'd remove CRLFs
198 */
199 if (stats->crlf) {
200 if (checksafe == SAFE_CRLF_WARN)
201 warning("CRLF will be replaced by LF in %s.\nThe file will have its original line endings in your working directory.", path);
202 else /* i.e. SAFE_CRLF_FAIL */
203 die("CRLF would be replaced by LF in %s.", path);
204 }
205 } else if (output_eol(crlf_action) == EOL_CRLF) {
206 /*
207 * CRLFs would be added by checkout:
208 * check if we have "naked" LFs
209 */
210 if (stats->lf != stats->crlf) {
211 if (checksafe == SAFE_CRLF_WARN)
212 warning("LF will be replaced by CRLF in %s.\nThe file will have its original line endings in your working directory.", path);
213 else /* i.e. SAFE_CRLF_FAIL */
214 die("LF would be replaced by CRLF in %s", path);
215 }
216 }
217 }
218
219 static int has_cr_in_index(const char *path)
220 {
221 unsigned long sz;
222 void *data;
223 int has_cr;
224
225 data = read_blob_data_from_cache(path, &sz);
226 if (!data)
227 return 0;
228 has_cr = memchr(data, '\r', sz) != NULL;
229 free(data);
230 return has_cr;
231 }
232
233 static int crlf_to_git(const char *path, const char *src, size_t len,
234 struct strbuf *buf,
235 enum crlf_action crlf_action, enum safe_crlf checksafe)
236 {
237 struct text_stat stats;
238 char *dst;
239
240 if (crlf_action == CRLF_BINARY ||
241 (src && !len))
242 return 0;
243
244 /*
245 * If we are doing a dry-run and have no source buffer, there is
246 * nothing to analyze; we must assume we would convert.
247 */
248 if (!buf && !src)
249 return 1;
250
251 gather_stats(src, len, &stats);
252
253 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
254 if (convert_is_binary(len, &stats))
255 return 0;
256
257 if (crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
258 /*
259 * If the file in the index has any CR in it, do not convert.
260 * This is the new safer autocrlf handling.
261 */
262 if (has_cr_in_index(path))
263 return 0;
264 }
265 }
266
267 check_safe_crlf(path, crlf_action, &stats, checksafe);
268
269 /* Optimization: No CR? Nothing to convert, regardless. */
270 if (!stats.cr)
271 return 0;
272
273 /*
274 * At this point all of our source analysis is done, and we are sure we
275 * would convert. If we are in dry-run mode, we can give an answer.
276 */
277 if (!buf)
278 return 1;
279
280 /* only grow if not in place */
281 if (strbuf_avail(buf) + buf->len < len)
282 strbuf_grow(buf, len - buf->len);
283 dst = buf->buf;
284 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
285 /*
286 * If we guessed, we already know we rejected a file with
287 * lone CR, and we can strip a CR without looking at what
288 * follow it.
289 */
290 do {
291 unsigned char c = *src++;
292 if (c != '\r')
293 *dst++ = c;
294 } while (--len);
295 } else {
296 do {
297 unsigned char c = *src++;
298 if (! (c == '\r' && (1 < len && *src == '\n')))
299 *dst++ = c;
300 } while (--len);
301 }
302 strbuf_setlen(buf, dst - buf->buf);
303 return 1;
304 }
305
306 static int crlf_to_worktree(const char *path, const char *src, size_t len,
307 struct strbuf *buf, enum crlf_action crlf_action)
308 {
309 char *to_free = NULL;
310 struct text_stat stats;
311
312 if (!len || output_eol(crlf_action) != EOL_CRLF)
313 return 0;
314
315 gather_stats(src, len, &stats);
316
317 /* No LF? Nothing to convert, regardless. */
318 if (!stats.lf)
319 return 0;
320
321 /* Was it already in CRLF format? */
322 if (stats.lf == stats.crlf)
323 return 0;
324
325 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
326 if (crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
327 /* If we have any CR or CRLF line endings, we do not touch it */
328 /* This is the new safer autocrlf-handling */
329 if (stats.cr > 0 || stats.crlf > 0)
330 return 0;
331 }
332
333 if (convert_is_binary(len, &stats))
334 return 0;
335 }
336
337 /* are we "faking" in place editing ? */
338 if (src == buf->buf)
339 to_free = strbuf_detach(buf, NULL);
340
341 strbuf_grow(buf, len + stats.lf - stats.crlf);
342 for (;;) {
343 const char *nl = memchr(src, '\n', len);
344 if (!nl)
345 break;
346 if (nl > src && nl[-1] == '\r') {
347 strbuf_add(buf, src, nl + 1 - src);
348 } else {
349 strbuf_add(buf, src, nl - src);
350 strbuf_addstr(buf, "\r\n");
351 }
352 len -= nl + 1 - src;
353 src = nl + 1;
354 }
355 strbuf_add(buf, src, len);
356
357 free(to_free);
358 return 1;
359 }
360
361 struct filter_params {
362 const char *src;
363 unsigned long size;
364 int fd;
365 const char *cmd;
366 const char *path;
367 };
368
369 static int filter_buffer_or_fd(int in, int out, void *data)
370 {
371 /*
372 * Spawn cmd and feed the buffer contents through its stdin.
373 */
374 struct child_process child_process = CHILD_PROCESS_INIT;
375 struct filter_params *params = (struct filter_params *)data;
376 int write_err, status;
377 const char *argv[] = { NULL, NULL };
378
379 /* apply % substitution to cmd */
380 struct strbuf cmd = STRBUF_INIT;
381 struct strbuf path = STRBUF_INIT;
382 struct strbuf_expand_dict_entry dict[] = {
383 { "f", NULL, },
384 { NULL, NULL, },
385 };
386
387 /* quote the path to preserve spaces, etc. */
388 sq_quote_buf(&path, params->path);
389 dict[0].value = path.buf;
390
391 /* expand all %f with the quoted path */
392 strbuf_expand(&cmd, params->cmd, strbuf_expand_dict_cb, &dict);
393 strbuf_release(&path);
394
395 argv[0] = cmd.buf;
396
397 child_process.argv = argv;
398 child_process.use_shell = 1;
399 child_process.in = -1;
400 child_process.out = out;
401
402 if (start_command(&child_process))
403 return error("cannot fork to run external filter %s", params->cmd);
404
405 sigchain_push(SIGPIPE, SIG_IGN);
406
407 if (params->src) {
408 write_err = (write_in_full(child_process.in,
409 params->src, params->size) < 0);
410 if (errno == EPIPE)
411 write_err = 0;
412 } else {
413 write_err = copy_fd(params->fd, child_process.in);
414 if (write_err == COPY_WRITE_ERROR && errno == EPIPE)
415 write_err = 0;
416 }
417
418 if (close(child_process.in))
419 write_err = 1;
420 if (write_err)
421 error("cannot feed the input to external filter %s", params->cmd);
422
423 sigchain_pop(SIGPIPE);
424
425 status = finish_command(&child_process);
426 if (status)
427 error("external filter %s failed %d", params->cmd, status);
428
429 strbuf_release(&cmd);
430 return (write_err || status);
431 }
432
433 static int apply_filter(const char *path, const char *src, size_t len, int fd,
434 struct strbuf *dst, const char *cmd)
435 {
436 /*
437 * Create a pipeline to have the command filter the buffer's
438 * contents.
439 *
440 * (child --> cmd) --> us
441 */
442 int ret = 1;
443 struct strbuf nbuf = STRBUF_INIT;
444 struct async async;
445 struct filter_params params;
446
447 if (!cmd)
448 return 0;
449
450 if (!dst)
451 return 1;
452
453 memset(&async, 0, sizeof(async));
454 async.proc = filter_buffer_or_fd;
455 async.data = &params;
456 async.out = -1;
457 params.src = src;
458 params.size = len;
459 params.fd = fd;
460 params.cmd = cmd;
461 params.path = path;
462
463 fflush(NULL);
464 if (start_async(&async))
465 return 0; /* error was already reported */
466
467 if (strbuf_read(&nbuf, async.out, len) < 0) {
468 error("read from external filter %s failed", cmd);
469 ret = 0;
470 }
471 if (close(async.out)) {
472 error("read from external filter %s failed", cmd);
473 ret = 0;
474 }
475 if (finish_async(&async)) {
476 error("external filter %s failed", cmd);
477 ret = 0;
478 }
479
480 if (ret) {
481 strbuf_swap(dst, &nbuf);
482 }
483 strbuf_release(&nbuf);
484 return ret;
485 }
486
487 static struct convert_driver {
488 const char *name;
489 struct convert_driver *next;
490 const char *smudge;
491 const char *clean;
492 int required;
493 } *user_convert, **user_convert_tail;
494
495 static int read_convert_config(const char *var, const char *value, void *cb)
496 {
497 const char *key, *name;
498 int namelen;
499 struct convert_driver *drv;
500
501 /*
502 * External conversion drivers are configured using
503 * "filter.<name>.variable".
504 */
505 if (parse_config_key(var, "filter", &name, &namelen, &key) < 0 || !name)
506 return 0;
507 for (drv = user_convert; drv; drv = drv->next)
508 if (!strncmp(drv->name, name, namelen) && !drv->name[namelen])
509 break;
510 if (!drv) {
511 drv = xcalloc(1, sizeof(struct convert_driver));
512 drv->name = xmemdupz(name, namelen);
513 *user_convert_tail = drv;
514 user_convert_tail = &(drv->next);
515 }
516
517 /*
518 * filter.<name>.smudge and filter.<name>.clean specifies
519 * the command line:
520 *
521 * command-line
522 *
523 * The command-line will not be interpolated in any way.
524 */
525
526 if (!strcmp("smudge", key))
527 return git_config_string(&drv->smudge, var, value);
528
529 if (!strcmp("clean", key))
530 return git_config_string(&drv->clean, var, value);
531
532 if (!strcmp("required", key)) {
533 drv->required = git_config_bool(var, value);
534 return 0;
535 }
536
537 return 0;
538 }
539
540 static int count_ident(const char *cp, unsigned long size)
541 {
542 /*
543 * "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$"
544 */
545 int cnt = 0;
546 char ch;
547
548 while (size) {
549 ch = *cp++;
550 size--;
551 if (ch != '$')
552 continue;
553 if (size < 3)
554 break;
555 if (memcmp("Id", cp, 2))
556 continue;
557 ch = cp[2];
558 cp += 3;
559 size -= 3;
560 if (ch == '$')
561 cnt++; /* $Id$ */
562 if (ch != ':')
563 continue;
564
565 /*
566 * "$Id: ... "; scan up to the closing dollar sign and discard.
567 */
568 while (size) {
569 ch = *cp++;
570 size--;
571 if (ch == '$') {
572 cnt++;
573 break;
574 }
575 if (ch == '\n')
576 break;
577 }
578 }
579 return cnt;
580 }
581
582 static int ident_to_git(const char *path, const char *src, size_t len,
583 struct strbuf *buf, int ident)
584 {
585 char *dst, *dollar;
586
587 if (!ident || (src && !count_ident(src, len)))
588 return 0;
589
590 if (!buf)
591 return 1;
592
593 /* only grow if not in place */
594 if (strbuf_avail(buf) + buf->len < len)
595 strbuf_grow(buf, len - buf->len);
596 dst = buf->buf;
597 for (;;) {
598 dollar = memchr(src, '$', len);
599 if (!dollar)
600 break;
601 memmove(dst, src, dollar + 1 - src);
602 dst += dollar + 1 - src;
603 len -= dollar + 1 - src;
604 src = dollar + 1;
605
606 if (len > 3 && !memcmp(src, "Id:", 3)) {
607 dollar = memchr(src + 3, '$', len - 3);
608 if (!dollar)
609 break;
610 if (memchr(src + 3, '\n', dollar - src - 3)) {
611 /* Line break before the next dollar. */
612 continue;
613 }
614
615 memcpy(dst, "Id$", 3);
616 dst += 3;
617 len -= dollar + 1 - src;
618 src = dollar + 1;
619 }
620 }
621 memmove(dst, src, len);
622 strbuf_setlen(buf, dst + len - buf->buf);
623 return 1;
624 }
625
626 static int ident_to_worktree(const char *path, const char *src, size_t len,
627 struct strbuf *buf, int ident)
628 {
629 unsigned char sha1[20];
630 char *to_free = NULL, *dollar, *spc;
631 int cnt;
632
633 if (!ident)
634 return 0;
635
636 cnt = count_ident(src, len);
637 if (!cnt)
638 return 0;
639
640 /* are we "faking" in place editing ? */
641 if (src == buf->buf)
642 to_free = strbuf_detach(buf, NULL);
643 hash_sha1_file(src, len, "blob", sha1);
644
645 strbuf_grow(buf, len + cnt * 43);
646 for (;;) {
647 /* step 1: run to the next '$' */
648 dollar = memchr(src, '$', len);
649 if (!dollar)
650 break;
651 strbuf_add(buf, src, dollar + 1 - src);
652 len -= dollar + 1 - src;
653 src = dollar + 1;
654
655 /* step 2: does it looks like a bit like Id:xxx$ or Id$ ? */
656 if (len < 3 || memcmp("Id", src, 2))
657 continue;
658
659 /* step 3: skip over Id$ or Id:xxxxx$ */
660 if (src[2] == '$') {
661 src += 3;
662 len -= 3;
663 } else if (src[2] == ':') {
664 /*
665 * It's possible that an expanded Id has crept its way into the
666 * repository, we cope with that by stripping the expansion out.
667 * This is probably not a good idea, since it will cause changes
668 * on checkout, which won't go away by stash, but let's keep it
669 * for git-style ids.
670 */
671 dollar = memchr(src + 3, '$', len - 3);
672 if (!dollar) {
673 /* incomplete keyword, no more '$', so just quit the loop */
674 break;
675 }
676
677 if (memchr(src + 3, '\n', dollar - src - 3)) {
678 /* Line break before the next dollar. */
679 continue;
680 }
681
682 spc = memchr(src + 4, ' ', dollar - src - 4);
683 if (spc && spc < dollar-1) {
684 /* There are spaces in unexpected places.
685 * This is probably an id from some other
686 * versioning system. Keep it for now.
687 */
688 continue;
689 }
690
691 len -= dollar + 1 - src;
692 src = dollar + 1;
693 } else {
694 /* it wasn't a "Id$" or "Id:xxxx$" */
695 continue;
696 }
697
698 /* step 4: substitute */
699 strbuf_addstr(buf, "Id: ");
700 strbuf_add(buf, sha1_to_hex(sha1), 40);
701 strbuf_addstr(buf, " $");
702 }
703 strbuf_add(buf, src, len);
704
705 free(to_free);
706 return 1;
707 }
708
709 static enum crlf_action git_path_check_crlf(struct git_attr_check *check)
710 {
711 const char *value = check->value;
712
713 if (ATTR_TRUE(value))
714 return text_eol_is_crlf() ? CRLF_TEXT_CRLF : CRLF_TEXT_INPUT;
715 else if (ATTR_FALSE(value))
716 return CRLF_BINARY;
717 else if (ATTR_UNSET(value))
718 ;
719 else if (!strcmp(value, "input"))
720 return CRLF_TEXT_INPUT;
721 else if (!strcmp(value, "auto"))
722 return CRLF_AUTO;
723 return CRLF_UNDEFINED;
724 }
725
726 static enum eol git_path_check_eol(struct git_attr_check *check)
727 {
728 const char *value = check->value;
729
730 if (ATTR_UNSET(value))
731 ;
732 else if (!strcmp(value, "lf"))
733 return EOL_LF;
734 else if (!strcmp(value, "crlf"))
735 return EOL_CRLF;
736 return EOL_UNSET;
737 }
738
739 static struct convert_driver *git_path_check_convert(struct git_attr_check *check)
740 {
741 const char *value = check->value;
742 struct convert_driver *drv;
743
744 if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
745 return NULL;
746 for (drv = user_convert; drv; drv = drv->next)
747 if (!strcmp(value, drv->name))
748 return drv;
749 return NULL;
750 }
751
752 static int git_path_check_ident(struct git_attr_check *check)
753 {
754 const char *value = check->value;
755
756 return !!ATTR_TRUE(value);
757 }
758
759 struct conv_attrs {
760 struct convert_driver *drv;
761 enum crlf_action attr_action; /* What attr says */
762 enum crlf_action crlf_action; /* When no attr is set, use core.autocrlf */
763 int ident;
764 };
765
766 static const char *conv_attr_name[] = {
767 "crlf", "ident", "filter", "eol", "text",
768 };
769 #define NUM_CONV_ATTRS ARRAY_SIZE(conv_attr_name)
770
771 static void convert_attrs(struct conv_attrs *ca, const char *path)
772 {
773 int i;
774 static struct git_attr_check ccheck[NUM_CONV_ATTRS];
775
776 if (!ccheck[0].attr) {
777 for (i = 0; i < NUM_CONV_ATTRS; i++)
778 ccheck[i].attr = git_attr(conv_attr_name[i]);
779 user_convert_tail = &user_convert;
780 git_config(read_convert_config, NULL);
781 }
782
783 if (!git_check_attr(path, NUM_CONV_ATTRS, ccheck)) {
784 enum eol eol_attr;
785 ca->crlf_action = git_path_check_crlf(ccheck + 4);
786 if (ca->crlf_action == CRLF_UNDEFINED)
787 ca->crlf_action = git_path_check_crlf(ccheck + 0);
788 ca->attr_action = ca->crlf_action;
789 ca->ident = git_path_check_ident(ccheck + 1);
790 ca->drv = git_path_check_convert(ccheck + 2);
791 if (ca->crlf_action == CRLF_BINARY)
792 return;
793 eol_attr = git_path_check_eol(ccheck + 3);
794 if (eol_attr == EOL_LF)
795 ca->crlf_action = CRLF_TEXT_INPUT;
796 else if (eol_attr == EOL_CRLF)
797 ca->crlf_action = CRLF_TEXT_CRLF;
798 } else {
799 ca->drv = NULL;
800 ca->crlf_action = CRLF_UNDEFINED;
801 ca->ident = 0;
802 }
803 if (ca->crlf_action == CRLF_TEXT)
804 ca->crlf_action = text_eol_is_crlf() ? CRLF_TEXT_CRLF : CRLF_TEXT_INPUT;
805 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_FALSE)
806 ca->crlf_action = CRLF_BINARY;
807 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_TRUE)
808 ca->crlf_action = CRLF_AUTO_CRLF;
809 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_INPUT)
810 ca->crlf_action = CRLF_AUTO_INPUT;
811 }
812
813 int would_convert_to_git_filter_fd(const char *path)
814 {
815 struct conv_attrs ca;
816
817 convert_attrs(&ca, path);
818 if (!ca.drv)
819 return 0;
820
821 /*
822 * Apply a filter to an fd only if the filter is required to succeed.
823 * We must die if the filter fails, because the original data before
824 * filtering is not available.
825 */
826 if (!ca.drv->required)
827 return 0;
828
829 return apply_filter(path, NULL, 0, -1, NULL, ca.drv->clean);
830 }
831
832 const char *get_convert_attr_ascii(const char *path)
833 {
834 struct conv_attrs ca;
835
836 convert_attrs(&ca, path);
837 switch (ca.attr_action) {
838 case CRLF_UNDEFINED:
839 return "";
840 case CRLF_BINARY:
841 return "-text";
842 case CRLF_TEXT:
843 return "text";
844 case CRLF_TEXT_INPUT:
845 return "text eol=lf";
846 case CRLF_TEXT_CRLF:
847 return "text eol=crlf";
848 case CRLF_AUTO:
849 return "text=auto";
850 case CRLF_AUTO_CRLF:
851 return "text=auto eol=crlf"; /* This is not supported yet */
852 case CRLF_AUTO_INPUT:
853 return "text=auto eol=lf"; /* This is not supported yet */
854 }
855 return "";
856 }
857
858 int convert_to_git(const char *path, const char *src, size_t len,
859 struct strbuf *dst, enum safe_crlf checksafe)
860 {
861 int ret = 0;
862 const char *filter = NULL;
863 int required = 0;
864 struct conv_attrs ca;
865
866 convert_attrs(&ca, path);
867 if (ca.drv) {
868 filter = ca.drv->clean;
869 required = ca.drv->required;
870 }
871
872 ret |= apply_filter(path, src, len, -1, dst, filter);
873 if (!ret && required)
874 die("%s: clean filter '%s' failed", path, ca.drv->name);
875
876 if (ret && dst) {
877 src = dst->buf;
878 len = dst->len;
879 }
880 ret |= crlf_to_git(path, src, len, dst, ca.crlf_action, checksafe);
881 if (ret && dst) {
882 src = dst->buf;
883 len = dst->len;
884 }
885 return ret | ident_to_git(path, src, len, dst, ca.ident);
886 }
887
888 void convert_to_git_filter_fd(const char *path, int fd, struct strbuf *dst,
889 enum safe_crlf checksafe)
890 {
891 struct conv_attrs ca;
892 convert_attrs(&ca, path);
893
894 assert(ca.drv);
895 assert(ca.drv->clean);
896
897 if (!apply_filter(path, NULL, 0, fd, dst, ca.drv->clean))
898 die("%s: clean filter '%s' failed", path, ca.drv->name);
899
900 crlf_to_git(path, dst->buf, dst->len, dst, ca.crlf_action, checksafe);
901 ident_to_git(path, dst->buf, dst->len, dst, ca.ident);
902 }
903
904 static int convert_to_working_tree_internal(const char *path, const char *src,
905 size_t len, struct strbuf *dst,
906 int normalizing)
907 {
908 int ret = 0, ret_filter = 0;
909 const char *filter = NULL;
910 int required = 0;
911 struct conv_attrs ca;
912
913 convert_attrs(&ca, path);
914 if (ca.drv) {
915 filter = ca.drv->smudge;
916 required = ca.drv->required;
917 }
918
919 ret |= ident_to_worktree(path, src, len, dst, ca.ident);
920 if (ret) {
921 src = dst->buf;
922 len = dst->len;
923 }
924 /*
925 * CRLF conversion can be skipped if normalizing, unless there
926 * is a smudge filter. The filter might expect CRLFs.
927 */
928 if (filter || !normalizing) {
929 ret |= crlf_to_worktree(path, src, len, dst, ca.crlf_action);
930 if (ret) {
931 src = dst->buf;
932 len = dst->len;
933 }
934 }
935
936 ret_filter = apply_filter(path, src, len, -1, dst, filter);
937 if (!ret_filter && required)
938 die("%s: smudge filter %s failed", path, ca.drv->name);
939
940 return ret | ret_filter;
941 }
942
943 int convert_to_working_tree(const char *path, const char *src, size_t len, struct strbuf *dst)
944 {
945 return convert_to_working_tree_internal(path, src, len, dst, 0);
946 }
947
948 int renormalize_buffer(const char *path, const char *src, size_t len, struct strbuf *dst)
949 {
950 int ret = convert_to_working_tree_internal(path, src, len, dst, 1);
951 if (ret) {
952 src = dst->buf;
953 len = dst->len;
954 }
955 return ret | convert_to_git(path, src, len, dst, SAFE_CRLF_FALSE);
956 }
957
958 /*****************************************************************
959 *
960 * Streaming conversion support
961 *
962 *****************************************************************/
963
964 typedef int (*filter_fn)(struct stream_filter *,
965 const char *input, size_t *isize_p,
966 char *output, size_t *osize_p);
967 typedef void (*free_fn)(struct stream_filter *);
968
969 struct stream_filter_vtbl {
970 filter_fn filter;
971 free_fn free;
972 };
973
974 struct stream_filter {
975 struct stream_filter_vtbl *vtbl;
976 };
977
978 static int null_filter_fn(struct stream_filter *filter,
979 const char *input, size_t *isize_p,
980 char *output, size_t *osize_p)
981 {
982 size_t count;
983
984 if (!input)
985 return 0; /* we do not keep any states */
986 count = *isize_p;
987 if (*osize_p < count)
988 count = *osize_p;
989 if (count) {
990 memmove(output, input, count);
991 *isize_p -= count;
992 *osize_p -= count;
993 }
994 return 0;
995 }
996
997 static void null_free_fn(struct stream_filter *filter)
998 {
999 ; /* nothing -- null instances are shared */
1000 }
1001
1002 static struct stream_filter_vtbl null_vtbl = {
1003 null_filter_fn,
1004 null_free_fn,
1005 };
1006
1007 static struct stream_filter null_filter_singleton = {
1008 &null_vtbl,
1009 };
1010
1011 int is_null_stream_filter(struct stream_filter *filter)
1012 {
1013 return filter == &null_filter_singleton;
1014 }
1015
1016
1017 /*
1018 * LF-to-CRLF filter
1019 */
1020
1021 struct lf_to_crlf_filter {
1022 struct stream_filter filter;
1023 unsigned has_held:1;
1024 char held;
1025 };
1026
1027 static int lf_to_crlf_filter_fn(struct stream_filter *filter,
1028 const char *input, size_t *isize_p,
1029 char *output, size_t *osize_p)
1030 {
1031 size_t count, o = 0;
1032 struct lf_to_crlf_filter *lf_to_crlf = (struct lf_to_crlf_filter *)filter;
1033
1034 /*
1035 * We may be holding onto the CR to see if it is followed by a
1036 * LF, in which case we would need to go to the main loop.
1037 * Otherwise, just emit it to the output stream.
1038 */
1039 if (lf_to_crlf->has_held && (lf_to_crlf->held != '\r' || !input)) {
1040 output[o++] = lf_to_crlf->held;
1041 lf_to_crlf->has_held = 0;
1042 }
1043
1044 /* We are told to drain */
1045 if (!input) {
1046 *osize_p -= o;
1047 return 0;
1048 }
1049
1050 count = *isize_p;
1051 if (count || lf_to_crlf->has_held) {
1052 size_t i;
1053 int was_cr = 0;
1054
1055 if (lf_to_crlf->has_held) {
1056 was_cr = 1;
1057 lf_to_crlf->has_held = 0;
1058 }
1059
1060 for (i = 0; o < *osize_p && i < count; i++) {
1061 char ch = input[i];
1062
1063 if (ch == '\n') {
1064 output[o++] = '\r';
1065 } else if (was_cr) {
1066 /*
1067 * Previous round saw CR and it is not followed
1068 * by a LF; emit the CR before processing the
1069 * current character.
1070 */
1071 output[o++] = '\r';
1072 }
1073
1074 /*
1075 * We may have consumed the last output slot,
1076 * in which case we need to break out of this
1077 * loop; hold the current character before
1078 * returning.
1079 */
1080 if (*osize_p <= o) {
1081 lf_to_crlf->has_held = 1;
1082 lf_to_crlf->held = ch;
1083 continue; /* break but increment i */
1084 }
1085
1086 if (ch == '\r') {
1087 was_cr = 1;
1088 continue;
1089 }
1090
1091 was_cr = 0;
1092 output[o++] = ch;
1093 }
1094
1095 *osize_p -= o;
1096 *isize_p -= i;
1097
1098 if (!lf_to_crlf->has_held && was_cr) {
1099 lf_to_crlf->has_held = 1;
1100 lf_to_crlf->held = '\r';
1101 }
1102 }
1103 return 0;
1104 }
1105
1106 static void lf_to_crlf_free_fn(struct stream_filter *filter)
1107 {
1108 free(filter);
1109 }
1110
1111 static struct stream_filter_vtbl lf_to_crlf_vtbl = {
1112 lf_to_crlf_filter_fn,
1113 lf_to_crlf_free_fn,
1114 };
1115
1116 static struct stream_filter *lf_to_crlf_filter(void)
1117 {
1118 struct lf_to_crlf_filter *lf_to_crlf = xcalloc(1, sizeof(*lf_to_crlf));
1119
1120 lf_to_crlf->filter.vtbl = &lf_to_crlf_vtbl;
1121 return (struct stream_filter *)lf_to_crlf;
1122 }
1123
1124 /*
1125 * Cascade filter
1126 */
1127 #define FILTER_BUFFER 1024
1128 struct cascade_filter {
1129 struct stream_filter filter;
1130 struct stream_filter *one;
1131 struct stream_filter *two;
1132 char buf[FILTER_BUFFER];
1133 int end, ptr;
1134 };
1135
1136 static int cascade_filter_fn(struct stream_filter *filter,
1137 const char *input, size_t *isize_p,
1138 char *output, size_t *osize_p)
1139 {
1140 struct cascade_filter *cas = (struct cascade_filter *) filter;
1141 size_t filled = 0;
1142 size_t sz = *osize_p;
1143 size_t to_feed, remaining;
1144
1145 /*
1146 * input -- (one) --> buf -- (two) --> output
1147 */
1148 while (filled < sz) {
1149 remaining = sz - filled;
1150
1151 /* do we already have something to feed two with? */
1152 if (cas->ptr < cas->end) {
1153 to_feed = cas->end - cas->ptr;
1154 if (stream_filter(cas->two,
1155 cas->buf + cas->ptr, &to_feed,
1156 output + filled, &remaining))
1157 return -1;
1158 cas->ptr += (cas->end - cas->ptr) - to_feed;
1159 filled = sz - remaining;
1160 continue;
1161 }
1162
1163 /* feed one from upstream and have it emit into our buffer */
1164 to_feed = input ? *isize_p : 0;
1165 if (input && !to_feed)
1166 break;
1167 remaining = sizeof(cas->buf);
1168 if (stream_filter(cas->one,
1169 input, &to_feed,
1170 cas->buf, &remaining))
1171 return -1;
1172 cas->end = sizeof(cas->buf) - remaining;
1173 cas->ptr = 0;
1174 if (input) {
1175 size_t fed = *isize_p - to_feed;
1176 *isize_p -= fed;
1177 input += fed;
1178 }
1179
1180 /* do we know that we drained one completely? */
1181 if (input || cas->end)
1182 continue;
1183
1184 /* tell two to drain; we have nothing more to give it */
1185 to_feed = 0;
1186 remaining = sz - filled;
1187 if (stream_filter(cas->two,
1188 NULL, &to_feed,
1189 output + filled, &remaining))
1190 return -1;
1191 if (remaining == (sz - filled))
1192 break; /* completely drained two */
1193 filled = sz - remaining;
1194 }
1195 *osize_p -= filled;
1196 return 0;
1197 }
1198
1199 static void cascade_free_fn(struct stream_filter *filter)
1200 {
1201 struct cascade_filter *cas = (struct cascade_filter *)filter;
1202 free_stream_filter(cas->one);
1203 free_stream_filter(cas->two);
1204 free(filter);
1205 }
1206
1207 static struct stream_filter_vtbl cascade_vtbl = {
1208 cascade_filter_fn,
1209 cascade_free_fn,
1210 };
1211
1212 static struct stream_filter *cascade_filter(struct stream_filter *one,
1213 struct stream_filter *two)
1214 {
1215 struct cascade_filter *cascade;
1216
1217 if (!one || is_null_stream_filter(one))
1218 return two;
1219 if (!two || is_null_stream_filter(two))
1220 return one;
1221
1222 cascade = xmalloc(sizeof(*cascade));
1223 cascade->one = one;
1224 cascade->two = two;
1225 cascade->end = cascade->ptr = 0;
1226 cascade->filter.vtbl = &cascade_vtbl;
1227 return (struct stream_filter *)cascade;
1228 }
1229
1230 /*
1231 * ident filter
1232 */
1233 #define IDENT_DRAINING (-1)
1234 #define IDENT_SKIPPING (-2)
1235 struct ident_filter {
1236 struct stream_filter filter;
1237 struct strbuf left;
1238 int state;
1239 char ident[45]; /* ": x40 $" */
1240 };
1241
1242 static int is_foreign_ident(const char *str)
1243 {
1244 int i;
1245
1246 if (!skip_prefix(str, "$Id: ", &str))
1247 return 0;
1248 for (i = 0; str[i]; i++) {
1249 if (isspace(str[i]) && str[i+1] != '$')
1250 return 1;
1251 }
1252 return 0;
1253 }
1254
1255 static void ident_drain(struct ident_filter *ident, char **output_p, size_t *osize_p)
1256 {
1257 size_t to_drain = ident->left.len;
1258
1259 if (*osize_p < to_drain)
1260 to_drain = *osize_p;
1261 if (to_drain) {
1262 memcpy(*output_p, ident->left.buf, to_drain);
1263 strbuf_remove(&ident->left, 0, to_drain);
1264 *output_p += to_drain;
1265 *osize_p -= to_drain;
1266 }
1267 if (!ident->left.len)
1268 ident->state = 0;
1269 }
1270
1271 static int ident_filter_fn(struct stream_filter *filter,
1272 const char *input, size_t *isize_p,
1273 char *output, size_t *osize_p)
1274 {
1275 struct ident_filter *ident = (struct ident_filter *)filter;
1276 static const char head[] = "$Id";
1277
1278 if (!input) {
1279 /* drain upon eof */
1280 switch (ident->state) {
1281 default:
1282 strbuf_add(&ident->left, head, ident->state);
1283 case IDENT_SKIPPING:
1284 /* fallthru */
1285 case IDENT_DRAINING:
1286 ident_drain(ident, &output, osize_p);
1287 }
1288 return 0;
1289 }
1290
1291 while (*isize_p || (ident->state == IDENT_DRAINING)) {
1292 int ch;
1293
1294 if (ident->state == IDENT_DRAINING) {
1295 ident_drain(ident, &output, osize_p);
1296 if (!*osize_p)
1297 break;
1298 continue;
1299 }
1300
1301 ch = *(input++);
1302 (*isize_p)--;
1303
1304 if (ident->state == IDENT_SKIPPING) {
1305 /*
1306 * Skipping until '$' or LF, but keeping them
1307 * in case it is a foreign ident.
1308 */
1309 strbuf_addch(&ident->left, ch);
1310 if (ch != '\n' && ch != '$')
1311 continue;
1312 if (ch == '$' && !is_foreign_ident(ident->left.buf)) {
1313 strbuf_setlen(&ident->left, sizeof(head) - 1);
1314 strbuf_addstr(&ident->left, ident->ident);
1315 }
1316 ident->state = IDENT_DRAINING;
1317 continue;
1318 }
1319
1320 if (ident->state < sizeof(head) &&
1321 head[ident->state] == ch) {
1322 ident->state++;
1323 continue;
1324 }
1325
1326 if (ident->state)
1327 strbuf_add(&ident->left, head, ident->state);
1328 if (ident->state == sizeof(head) - 1) {
1329 if (ch != ':' && ch != '$') {
1330 strbuf_addch(&ident->left, ch);
1331 ident->state = 0;
1332 continue;
1333 }
1334
1335 if (ch == ':') {
1336 strbuf_addch(&ident->left, ch);
1337 ident->state = IDENT_SKIPPING;
1338 } else {
1339 strbuf_addstr(&ident->left, ident->ident);
1340 ident->state = IDENT_DRAINING;
1341 }
1342 continue;
1343 }
1344
1345 strbuf_addch(&ident->left, ch);
1346 ident->state = IDENT_DRAINING;
1347 }
1348 return 0;
1349 }
1350
1351 static void ident_free_fn(struct stream_filter *filter)
1352 {
1353 struct ident_filter *ident = (struct ident_filter *)filter;
1354 strbuf_release(&ident->left);
1355 free(filter);
1356 }
1357
1358 static struct stream_filter_vtbl ident_vtbl = {
1359 ident_filter_fn,
1360 ident_free_fn,
1361 };
1362
1363 static struct stream_filter *ident_filter(const unsigned char *sha1)
1364 {
1365 struct ident_filter *ident = xmalloc(sizeof(*ident));
1366
1367 xsnprintf(ident->ident, sizeof(ident->ident),
1368 ": %s $", sha1_to_hex(sha1));
1369 strbuf_init(&ident->left, 0);
1370 ident->filter.vtbl = &ident_vtbl;
1371 ident->state = 0;
1372 return (struct stream_filter *)ident;
1373 }
1374
1375 /*
1376 * Return an appropriately constructed filter for the path, or NULL if
1377 * the contents cannot be filtered without reading the whole thing
1378 * in-core.
1379 *
1380 * Note that you would be crazy to set CRLF, smuge/clean or ident to a
1381 * large binary blob you would want us not to slurp into the memory!
1382 */
1383 struct stream_filter *get_stream_filter(const char *path, const unsigned char *sha1)
1384 {
1385 struct conv_attrs ca;
1386 enum crlf_action crlf_action;
1387 struct stream_filter *filter = NULL;
1388
1389 convert_attrs(&ca, path);
1390
1391 if (ca.drv && (ca.drv->smudge || ca.drv->clean))
1392 return filter;
1393
1394 if (ca.ident)
1395 filter = ident_filter(sha1);
1396
1397 crlf_action = ca.crlf_action;
1398
1399 if ((crlf_action == CRLF_BINARY) ||
1400 crlf_action == CRLF_AUTO_INPUT ||
1401 (crlf_action == CRLF_TEXT_INPUT))
1402 filter = cascade_filter(filter, &null_filter_singleton);
1403
1404 else if (output_eol(crlf_action) == EOL_CRLF &&
1405 !(crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_CRLF))
1406 filter = cascade_filter(filter, lf_to_crlf_filter());
1407
1408 return filter;
1409 }
1410
1411 void free_stream_filter(struct stream_filter *filter)
1412 {
1413 filter->vtbl->free(filter);
1414 }
1415
1416 int stream_filter(struct stream_filter *filter,
1417 const char *input, size_t *isize_p,
1418 char *output, size_t *osize_p)
1419 {
1420 return filter->vtbl->filter(filter, input, isize_p, output, osize_p);
1421 }