lf_to_crlf_filter(): tell the caller we added "\n" when draining
[git/git.git] / convert.c
CommitLineData
6c510bee 1#include "cache.h"
35ebfd6a 2#include "attr.h"
3fed15f5 3#include "run-command.h"
a2b665de 4#include "quote.h"
35ebfd6a 5
6c510bee
LT
6/*
7 * convert.c - convert a file when checking it out and checking it in.
8 *
9 * This should use the pathname to decide on whether it wants to do some
10 * more interesting conversions (automatic gzip/unzip, general format
11 * conversions etc etc), but by default it just does automatic CRLF<->LF
942e7747 12 * translation when the "text" attribute or "auto_crlf" option is set.
6c510bee
LT
13 */
14
c61dcff9 15enum crlf_action {
fd6cce9e
EB
16 CRLF_GUESS = -1,
17 CRLF_BINARY = 0,
18 CRLF_TEXT,
19 CRLF_INPUT,
20 CRLF_CRLF,
c9b6782a 21 CRLF_AUTO
fd6cce9e 22};
163b9591 23
6c510bee 24struct text_stat {
28624193
DP
25 /* NUL, CR, LF and CRLF counts */
26 unsigned nul, cr, lf, crlf;
6c510bee
LT
27
28 /* These are just approximations! */
29 unsigned printable, nonprintable;
30};
31
32static void gather_stats(const char *buf, unsigned long size, struct text_stat *stats)
33{
34 unsigned long i;
35
36 memset(stats, 0, sizeof(*stats));
37
38 for (i = 0; i < size; i++) {
39 unsigned char c = buf[i];
40 if (c == '\r') {
41 stats->cr++;
42 if (i+1 < size && buf[i+1] == '\n')
43 stats->crlf++;
44 continue;
45 }
46 if (c == '\n') {
47 stats->lf++;
48 continue;
49 }
50 if (c == 127)
51 /* DEL */
52 stats->nonprintable++;
53 else if (c < 32) {
54 switch (c) {
55 /* BS, HT, ESC and FF */
56 case '\b': case '\t': case '\033': case '\014':
57 stats->printable++;
58 break;
28624193
DP
59 case 0:
60 stats->nul++;
61 /* fall through */
6c510bee
LT
62 default:
63 stats->nonprintable++;
64 }
65 }
66 else
67 stats->printable++;
68 }
f9dd4bf4
DK
69
70 /* If file ends with EOF then don't count this EOF as non-printable. */
71 if (size >= 1 && buf[size-1] == '\032')
72 stats->nonprintable--;
6c510bee
LT
73}
74
75/*
76 * The same heuristics as diff.c::mmfile_is_binary()
77 */
78static int is_binary(unsigned long size, struct text_stat *stats)
79{
80
28624193
DP
81 if (stats->nul)
82 return 1;
6c510bee
LT
83 if ((stats->printable >> 7) < stats->nonprintable)
84 return 1;
85 /*
86 * Other heuristics? Average line length might be relevant,
87 * as might LF vs CR vs CRLF counts..
88 *
89 * NOTE! It might be normal to have a low ratio of CRLF to LF
90 * (somebody starts with a LF-only file and edits it with an editor
91 * that adds CRLF only to lines that are added..). But do we
92 * want to support CR-only? Probably not.
93 */
94 return 0;
95}
96
c61dcff9 97static enum eol output_eol(enum crlf_action crlf_action)
f217f0e8 98{
c61dcff9 99 switch (crlf_action) {
942e7747
EB
100 case CRLF_BINARY:
101 return EOL_UNSET;
102 case CRLF_CRLF:
103 return EOL_CRLF;
104 case CRLF_INPUT:
105 return EOL_LF;
106 case CRLF_GUESS:
107 if (!auto_crlf)
108 return EOL_UNSET;
109 /* fall through */
110 case CRLF_TEXT:
111 case CRLF_AUTO:
112 if (auto_crlf == AUTO_CRLF_TRUE)
113 return EOL_CRLF;
114 else if (auto_crlf == AUTO_CRLF_INPUT)
115 return EOL_LF;
ec70f52f 116 else if (core_eol == EOL_UNSET)
942e7747
EB
117 return EOL_NATIVE;
118 }
ec70f52f 119 return core_eol;
942e7747
EB
120}
121
c61dcff9 122static void check_safe_crlf(const char *path, enum crlf_action crlf_action,
21e5ad50
SP
123 struct text_stat *stats, enum safe_crlf checksafe)
124{
125 if (!checksafe)
126 return;
127
c61dcff9 128 if (output_eol(crlf_action) == EOL_LF) {
21e5ad50
SP
129 /*
130 * CRLFs would not be restored by checkout:
131 * check if we'd remove CRLFs
132 */
133 if (stats->crlf) {
134 if (checksafe == SAFE_CRLF_WARN)
942e7747 135 warning("CRLF will be replaced by LF in %s.\nThe file will have its original line endings in your working directory.", path);
21e5ad50
SP
136 else /* i.e. SAFE_CRLF_FAIL */
137 die("CRLF would be replaced by LF in %s.", path);
138 }
c61dcff9 139 } else if (output_eol(crlf_action) == EOL_CRLF) {
21e5ad50
SP
140 /*
141 * CRLFs would be added by checkout:
142 * check if we have "naked" LFs
143 */
144 if (stats->lf != stats->crlf) {
145 if (checksafe == SAFE_CRLF_WARN)
942e7747 146 warning("LF will be replaced by CRLF in %s.\nThe file will have its original line endings in your working directory.", path);
21e5ad50
SP
147 else /* i.e. SAFE_CRLF_FAIL */
148 die("LF would be replaced by CRLF in %s", path);
149 }
150 }
151}
152
c4805393
FAG
153static int has_cr_in_index(const char *path)
154{
155 int pos, len;
156 unsigned long sz;
157 enum object_type type;
158 void *data;
159 int has_cr;
160 struct index_state *istate = &the_index;
161
162 len = strlen(path);
163 pos = index_name_pos(istate, path, len);
164 if (pos < 0) {
165 /*
166 * We might be in the middle of a merge, in which
167 * case we would read stage #2 (ours).
168 */
169 int i;
170 for (i = -pos - 1;
171 (pos < 0 && i < istate->cache_nr &&
172 !strcmp(istate->cache[i]->name, path));
173 i++)
174 if (ce_stage(istate->cache[i]) == 2)
175 pos = i;
176 }
177 if (pos < 0)
178 return 0;
179 data = read_sha1_file(istate->cache[pos]->sha1, &type, &sz);
180 if (!data || type != OBJ_BLOB) {
181 free(data);
182 return 0;
183 }
184
185 has_cr = memchr(data, '\r', sz) != NULL;
186 free(data);
187 return has_cr;
188}
189
5ecd293d 190static int crlf_to_git(const char *path, const char *src, size_t len,
c61dcff9
JH
191 struct strbuf *buf,
192 enum crlf_action crlf_action, enum safe_crlf checksafe)
6c510bee 193{
6c510bee 194 struct text_stat stats;
5ecd293d 195 char *dst;
6c510bee 196
c61dcff9
JH
197 if (crlf_action == CRLF_BINARY ||
198 (crlf_action == CRLF_GUESS && auto_crlf == AUTO_CRLF_FALSE) || !len)
5ecd293d 199 return 0;
6c510bee 200
5ecd293d 201 gather_stats(src, len, &stats);
6c510bee 202
c61dcff9 203 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_GUESS) {
201ac8ef
JH
204 /*
205 * We're currently not going to even try to convert stuff
206 * that has bare CR characters. Does anybody do that crazy
207 * stuff?
208 */
209 if (stats.cr != stats.crlf)
5ecd293d 210 return 0;
201ac8ef
JH
211
212 /*
213 * And add some heuristics for binary vs text, of course...
214 */
5ecd293d
PH
215 if (is_binary(len, &stats))
216 return 0;
c4805393 217
c61dcff9 218 if (crlf_action == CRLF_GUESS) {
fd6cce9e
EB
219 /*
220 * If the file in the index has any CR in it, do not convert.
221 * This is the new safer autocrlf handling.
222 */
223 if (has_cr_in_index(path))
224 return 0;
225 }
201ac8ef 226 }
6c510bee 227
c61dcff9 228 check_safe_crlf(path, crlf_action, &stats, checksafe);
21e5ad50
SP
229
230 /* Optimization: No CR? Nothing to convert, regardless. */
231 if (!stats.cr)
232 return 0;
233
90d16ec0
PH
234 /* only grow if not in place */
235 if (strbuf_avail(buf) + buf->len < len)
236 strbuf_grow(buf, len - buf->len);
5ecd293d 237 dst = buf->buf;
c61dcff9 238 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_GUESS) {
163b9591
JH
239 /*
240 * If we guessed, we already know we rejected a file with
241 * lone CR, and we can strip a CR without looking at what
242 * follow it.
243 */
201ac8ef 244 do {
ac78e548 245 unsigned char c = *src++;
201ac8ef 246 if (c != '\r')
ac78e548 247 *dst++ = c;
5ecd293d 248 } while (--len);
201ac8ef
JH
249 } else {
250 do {
ac78e548 251 unsigned char c = *src++;
5ecd293d 252 if (! (c == '\r' && (1 < len && *src == '\n')))
ac78e548 253 *dst++ = c;
5ecd293d 254 } while (--len);
201ac8ef 255 }
5ecd293d
PH
256 strbuf_setlen(buf, dst - buf->buf);
257 return 1;
6c510bee
LT
258}
259
5ecd293d 260static int crlf_to_worktree(const char *path, const char *src, size_t len,
c61dcff9 261 struct strbuf *buf, enum crlf_action crlf_action)
6c510bee 262{
5ecd293d 263 char *to_free = NULL;
6c510bee 264 struct text_stat stats;
6c510bee 265
c61dcff9 266 if (!len || output_eol(crlf_action) != EOL_CRLF)
5ecd293d 267 return 0;
6c510bee 268
5ecd293d 269 gather_stats(src, len, &stats);
6c510bee
LT
270
271 /* No LF? Nothing to convert, regardless. */
272 if (!stats.lf)
5ecd293d 273 return 0;
6c510bee
LT
274
275 /* Was it already in CRLF format? */
276 if (stats.lf == stats.crlf)
5ecd293d 277 return 0;
6c510bee 278
c61dcff9
JH
279 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_GUESS) {
280 if (crlf_action == CRLF_GUESS) {
fd6cce9e
EB
281 /* If we have any CR or CRLF line endings, we do not touch it */
282 /* This is the new safer autocrlf-handling */
283 if (stats.cr > 0 || stats.crlf > 0)
284 return 0;
285 }
c4805393 286
201ac8ef
JH
287 /* If we have any bare CR characters, we're not going to touch it */
288 if (stats.cr != stats.crlf)
5ecd293d 289 return 0;
6c510bee 290
5ecd293d
PH
291 if (is_binary(len, &stats))
292 return 0;
201ac8ef 293 }
6c510bee 294
5ecd293d
PH
295 /* are we "faking" in place editing ? */
296 if (src == buf->buf)
b315c5c0 297 to_free = strbuf_detach(buf, NULL);
5ecd293d
PH
298
299 strbuf_grow(buf, len + stats.lf - stats.crlf);
300 for (;;) {
301 const char *nl = memchr(src, '\n', len);
302 if (!nl)
303 break;
304 if (nl > src && nl[-1] == '\r') {
305 strbuf_add(buf, src, nl + 1 - src);
306 } else {
307 strbuf_add(buf, src, nl - src);
308 strbuf_addstr(buf, "\r\n");
309 }
310 len -= nl + 1 - src;
311 src = nl + 1;
312 }
313 strbuf_add(buf, src, len);
314
315 free(to_free);
316 return 1;
6c510bee 317}
35ebfd6a 318
546bb582
JS
319struct filter_params {
320 const char *src;
321 unsigned long size;
322 const char *cmd;
a2b665de 323 const char *path;
546bb582
JS
324};
325
ae6a5609 326static int filter_buffer(int in, int out, void *data)
aa4ed402
JH
327{
328 /*
329 * Spawn cmd and feed the buffer contents through its stdin.
330 */
331 struct child_process child_process;
546bb582 332 struct filter_params *params = (struct filter_params *)data;
aa4ed402 333 int write_err, status;
66dbfd55
GV
334 const char *argv[] = { NULL, NULL };
335
a2b665de
PW
336 /* apply % substitution to cmd */
337 struct strbuf cmd = STRBUF_INIT;
338 struct strbuf path = STRBUF_INIT;
339 struct strbuf_expand_dict_entry dict[] = {
340 { "f", NULL, },
341 { NULL, NULL, },
342 };
343
344 /* quote the path to preserve spaces, etc. */
345 sq_quote_buf(&path, params->path);
346 dict[0].value = path.buf;
347
348 /* expand all %f with the quoted path */
349 strbuf_expand(&cmd, params->cmd, strbuf_expand_dict_cb, &dict);
350 strbuf_release(&path);
351
352 argv[0] = cmd.buf;
aa4ed402
JH
353
354 memset(&child_process, 0, sizeof(child_process));
dc1bfdcd 355 child_process.argv = argv;
ac0ba18d 356 child_process.use_shell = 1;
dc1bfdcd 357 child_process.in = -1;
ae6a5609 358 child_process.out = out;
aa4ed402 359
dc1bfdcd 360 if (start_command(&child_process))
546bb582 361 return error("cannot fork to run external filter %s", params->cmd);
aa4ed402 362
546bb582 363 write_err = (write_in_full(child_process.in, params->src, params->size) < 0);
dc1bfdcd 364 if (close(child_process.in))
aa4ed402
JH
365 write_err = 1;
366 if (write_err)
546bb582 367 error("cannot feed the input to external filter %s", params->cmd);
aa4ed402
JH
368
369 status = finish_command(&child_process);
370 if (status)
5709e036 371 error("external filter %s failed %d", params->cmd, status);
a2b665de
PW
372
373 strbuf_release(&cmd);
aa4ed402
JH
374 return (write_err || status);
375}
376
5ecd293d
PH
377static int apply_filter(const char *path, const char *src, size_t len,
378 struct strbuf *dst, const char *cmd)
aa4ed402
JH
379{
380 /*
381 * Create a pipeline to have the command filter the buffer's
382 * contents.
383 *
384 * (child --> cmd) --> us
385 */
546bb582 386 int ret = 1;
f285a2d7 387 struct strbuf nbuf = STRBUF_INIT;
546bb582
JS
388 struct async async;
389 struct filter_params params;
aa4ed402
JH
390
391 if (!cmd)
5ecd293d 392 return 0;
aa4ed402 393
546bb582
JS
394 memset(&async, 0, sizeof(async));
395 async.proc = filter_buffer;
396 async.data = &params;
ae6a5609 397 async.out = -1;
546bb582
JS
398 params.src = src;
399 params.size = len;
400 params.cmd = cmd;
a2b665de 401 params.path = path;
aa4ed402
JH
402
403 fflush(NULL);
546bb582
JS
404 if (start_async(&async))
405 return 0; /* error was already reported */
aa4ed402 406
546bb582 407 if (strbuf_read(&nbuf, async.out, len) < 0) {
5ecd293d
PH
408 error("read from external filter %s failed", cmd);
409 ret = 0;
aa4ed402 410 }
546bb582 411 if (close(async.out)) {
90d16ec0 412 error("read from external filter %s failed", cmd);
5ecd293d 413 ret = 0;
aa4ed402 414 }
546bb582
JS
415 if (finish_async(&async)) {
416 error("external filter %s failed", cmd);
5ecd293d 417 ret = 0;
aa4ed402
JH
418 }
419
5ecd293d 420 if (ret) {
90d16ec0 421 strbuf_swap(dst, &nbuf);
5ecd293d 422 }
90d16ec0 423 strbuf_release(&nbuf);
5ecd293d 424 return ret;
aa4ed402
JH
425}
426
427static struct convert_driver {
428 const char *name;
429 struct convert_driver *next;
cd8be6c9
BH
430 const char *smudge;
431 const char *clean;
aa4ed402
JH
432} *user_convert, **user_convert_tail;
433
ef90d6d4 434static int read_convert_config(const char *var, const char *value, void *cb)
aa4ed402
JH
435{
436 const char *ep, *name;
437 int namelen;
438 struct convert_driver *drv;
439
440 /*
441 * External conversion drivers are configured using
442 * "filter.<name>.variable".
443 */
444 if (prefixcmp(var, "filter.") || (ep = strrchr(var, '.')) == var + 6)
445 return 0;
446 name = var + 7;
447 namelen = ep - name;
448 for (drv = user_convert; drv; drv = drv->next)
449 if (!strncmp(drv->name, name, namelen) && !drv->name[namelen])
450 break;
451 if (!drv) {
aa4ed402 452 drv = xcalloc(1, sizeof(struct convert_driver));
182af834 453 drv->name = xmemdupz(name, namelen);
aa4ed402
JH
454 *user_convert_tail = drv;
455 user_convert_tail = &(drv->next);
456 }
457
458 ep++;
459
460 /*
461 * filter.<name>.smudge and filter.<name>.clean specifies
462 * the command line:
463 *
464 * command-line
465 *
466 * The command-line will not be interpolated in any way.
467 */
468
cd8be6c9
BH
469 if (!strcmp("smudge", ep))
470 return git_config_string(&drv->smudge, var, value);
471
472 if (!strcmp("clean", ep))
473 return git_config_string(&drv->clean, var, value);
aa4ed402 474
aa4ed402
JH
475 return 0;
476}
477
3fed15f5
JH
478static int count_ident(const char *cp, unsigned long size)
479{
480 /*
af9b54bb 481 * "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$"
3fed15f5
JH
482 */
483 int cnt = 0;
484 char ch;
485
486 while (size) {
487 ch = *cp++;
488 size--;
489 if (ch != '$')
490 continue;
af9b54bb 491 if (size < 3)
3fed15f5 492 break;
af9b54bb 493 if (memcmp("Id", cp, 2))
3fed15f5 494 continue;
af9b54bb
AP
495 ch = cp[2];
496 cp += 3;
497 size -= 3;
3fed15f5 498 if (ch == '$')
af9b54bb 499 cnt++; /* $Id$ */
3fed15f5
JH
500 if (ch != ':')
501 continue;
502
503 /*
af9b54bb 504 * "$Id: ... "; scan up to the closing dollar sign and discard.
3fed15f5
JH
505 */
506 while (size) {
507 ch = *cp++;
508 size--;
509 if (ch == '$') {
510 cnt++;
511 break;
512 }
a9f3049f
HG
513 if (ch == '\n')
514 break;
3fed15f5
JH
515 }
516 }
517 return cnt;
518}
519
5ecd293d
PH
520static int ident_to_git(const char *path, const char *src, size_t len,
521 struct strbuf *buf, int ident)
3fed15f5 522{
5ecd293d 523 char *dst, *dollar;
3fed15f5 524
5ecd293d
PH
525 if (!ident || !count_ident(src, len))
526 return 0;
527
90d16ec0
PH
528 /* only grow if not in place */
529 if (strbuf_avail(buf) + buf->len < len)
530 strbuf_grow(buf, len - buf->len);
5ecd293d
PH
531 dst = buf->buf;
532 for (;;) {
533 dollar = memchr(src, '$', len);
534 if (!dollar)
535 break;
77321184 536 memmove(dst, src, dollar + 1 - src);
5ecd293d
PH
537 dst += dollar + 1 - src;
538 len -= dollar + 1 - src;
539 src = dollar + 1;
540
541 if (len > 3 && !memcmp(src, "Id:", 3)) {
542 dollar = memchr(src + 3, '$', len - 3);
543 if (!dollar)
544 break;
a9f3049f
HG
545 if (memchr(src + 3, '\n', dollar - src - 3)) {
546 /* Line break before the next dollar. */
547 continue;
548 }
549
af9b54bb
AP
550 memcpy(dst, "Id$", 3);
551 dst += 3;
5ecd293d
PH
552 len -= dollar + 1 - src;
553 src = dollar + 1;
3fed15f5
JH
554 }
555 }
77321184 556 memmove(dst, src, len);
5ecd293d
PH
557 strbuf_setlen(buf, dst + len - buf->buf);
558 return 1;
3fed15f5
JH
559}
560
5ecd293d
PH
561static int ident_to_worktree(const char *path, const char *src, size_t len,
562 struct strbuf *buf, int ident)
3fed15f5 563{
3fed15f5 564 unsigned char sha1[20];
07814d90 565 char *to_free = NULL, *dollar, *spc;
5ecd293d 566 int cnt;
3fed15f5
JH
567
568 if (!ident)
5ecd293d 569 return 0;
3fed15f5 570
5ecd293d 571 cnt = count_ident(src, len);
3fed15f5 572 if (!cnt)
5ecd293d 573 return 0;
3fed15f5 574
5ecd293d
PH
575 /* are we "faking" in place editing ? */
576 if (src == buf->buf)
b315c5c0 577 to_free = strbuf_detach(buf, NULL);
5ecd293d 578 hash_sha1_file(src, len, "blob", sha1);
3fed15f5 579
5ecd293d
PH
580 strbuf_grow(buf, len + cnt * 43);
581 for (;;) {
582 /* step 1: run to the next '$' */
583 dollar = memchr(src, '$', len);
584 if (!dollar)
585 break;
586 strbuf_add(buf, src, dollar + 1 - src);
587 len -= dollar + 1 - src;
588 src = dollar + 1;
c23290d5 589
5ecd293d
PH
590 /* step 2: does it looks like a bit like Id:xxx$ or Id$ ? */
591 if (len < 3 || memcmp("Id", src, 2))
3fed15f5
JH
592 continue;
593
5ecd293d
PH
594 /* step 3: skip over Id$ or Id:xxxxx$ */
595 if (src[2] == '$') {
596 src += 3;
597 len -= 3;
598 } else if (src[2] == ':') {
599 /*
600 * It's possible that an expanded Id has crept its way into the
07814d90
HG
601 * repository, we cope with that by stripping the expansion out.
602 * This is probably not a good idea, since it will cause changes
603 * on checkout, which won't go away by stash, but let's keep it
604 * for git-style ids.
5ecd293d
PH
605 */
606 dollar = memchr(src + 3, '$', len - 3);
607 if (!dollar) {
608 /* incomplete keyword, no more '$', so just quit the loop */
609 break;
610 }
c23290d5 611
a9f3049f
HG
612 if (memchr(src + 3, '\n', dollar - src - 3)) {
613 /* Line break before the next dollar. */
614 continue;
615 }
616
07814d90
HG
617 spc = memchr(src + 4, ' ', dollar - src - 4);
618 if (spc && spc < dollar-1) {
619 /* There are spaces in unexpected places.
620 * This is probably an id from some other
621 * versioning system. Keep it for now.
622 */
623 continue;
624 }
625
5ecd293d
PH
626 len -= dollar + 1 - src;
627 src = dollar + 1;
628 } else {
629 /* it wasn't a "Id$" or "Id:xxxx$" */
630 continue;
631 }
c23290d5 632
5ecd293d
PH
633 /* step 4: substitute */
634 strbuf_addstr(buf, "Id: ");
635 strbuf_add(buf, sha1_to_hex(sha1), 40);
636 strbuf_addstr(buf, " $");
3fed15f5 637 }
5ecd293d 638 strbuf_add(buf, src, len);
3fed15f5 639
5ecd293d
PH
640 free(to_free);
641 return 1;
35ebfd6a
JH
642}
643
6073ee85 644static int git_path_check_crlf(const char *path, struct git_attr_check *check)
35ebfd6a 645{
6073ee85
JH
646 const char *value = check->value;
647
648 if (ATTR_TRUE(value))
649 return CRLF_TEXT;
650 else if (ATTR_FALSE(value))
651 return CRLF_BINARY;
652 else if (ATTR_UNSET(value))
653 ;
654 else if (!strcmp(value, "input"))
655 return CRLF_INPUT;
fd6cce9e
EB
656 else if (!strcmp(value, "auto"))
657 return CRLF_AUTO;
163b9591 658 return CRLF_GUESS;
35ebfd6a
JH
659}
660
fd6cce9e
EB
661static int git_path_check_eol(const char *path, struct git_attr_check *check)
662{
663 const char *value = check->value;
664
665 if (ATTR_UNSET(value))
666 ;
667 else if (!strcmp(value, "lf"))
668 return EOL_LF;
669 else if (!strcmp(value, "crlf"))
670 return EOL_CRLF;
671 return EOL_UNSET;
672}
673
aa4ed402
JH
674static struct convert_driver *git_path_check_convert(const char *path,
675 struct git_attr_check *check)
676{
677 const char *value = check->value;
678 struct convert_driver *drv;
679
680 if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
681 return NULL;
682 for (drv = user_convert; drv; drv = drv->next)
683 if (!strcmp(value, drv->name))
684 return drv;
685 return NULL;
686}
687
3fed15f5
JH
688static int git_path_check_ident(const char *path, struct git_attr_check *check)
689{
690 const char *value = check->value;
691
692 return !!ATTR_TRUE(value);
693}
694
c61dcff9 695static enum crlf_action input_crlf_action(enum crlf_action text_attr, enum eol eol_attr)
f217f0e8 696{
5ec3e670 697 if (text_attr == CRLF_BINARY)
fd6cce9e
EB
698 return CRLF_BINARY;
699 if (eol_attr == EOL_LF)
700 return CRLF_INPUT;
701 if (eol_attr == EOL_CRLF)
702 return CRLF_CRLF;
5ec3e670 703 return text_attr;
fd6cce9e
EB
704}
705
3bfba20d
JH
706struct conv_attrs {
707 struct convert_driver *drv;
708 enum crlf_action crlf_action;
709 enum eol eol_attr;
710 int ident;
711};
712
83295964
JH
713static const char *conv_attr_name[] = {
714 "crlf", "ident", "filter", "eol", "text",
715};
716#define NUM_CONV_ATTRS ARRAY_SIZE(conv_attr_name)
717
3bfba20d 718static void convert_attrs(struct conv_attrs *ca, const char *path)
83295964
JH
719{
720 int i;
721 static struct git_attr_check ccheck[NUM_CONV_ATTRS];
722
723 if (!ccheck[0].attr) {
724 for (i = 0; i < NUM_CONV_ATTRS; i++)
725 ccheck[i].attr = git_attr(conv_attr_name[i]);
726 user_convert_tail = &user_convert;
727 git_config(read_convert_config, NULL);
728 }
3bfba20d 729
d932f4eb 730 if (!git_check_attr(path, NUM_CONV_ATTRS, ccheck)) {
3bfba20d
JH
731 ca->crlf_action = git_path_check_crlf(path, ccheck + 4);
732 if (ca->crlf_action == CRLF_GUESS)
733 ca->crlf_action = git_path_check_crlf(path, ccheck + 0);
734 ca->ident = git_path_check_ident(path, ccheck + 1);
735 ca->drv = git_path_check_convert(path, ccheck + 2);
736 ca->eol_attr = git_path_check_eol(path, ccheck + 3);
737 } else {
738 ca->drv = NULL;
739 ca->crlf_action = CRLF_GUESS;
740 ca->eol_attr = EOL_UNSET;
741 ca->ident = 0;
742 }
83295964
JH
743}
744
21e5ad50
SP
745int convert_to_git(const char *path, const char *src, size_t len,
746 struct strbuf *dst, enum safe_crlf checksafe)
35ebfd6a 747{
3bfba20d 748 int ret = 0;
cd8be6c9 749 const char *filter = NULL;
3bfba20d 750 struct conv_attrs ca;
6073ee85 751
3bfba20d
JH
752 convert_attrs(&ca, path);
753 if (ca.drv)
754 filter = ca.drv->clean;
3fed15f5 755
5ecd293d
PH
756 ret |= apply_filter(path, src, len, dst, filter);
757 if (ret) {
758 src = dst->buf;
759 len = dst->len;
aa4ed402 760 }
3bfba20d
JH
761 ca.crlf_action = input_crlf_action(ca.crlf_action, ca.eol_attr);
762 ret |= crlf_to_git(path, src, len, dst, ca.crlf_action, checksafe);
5ecd293d
PH
763 if (ret) {
764 src = dst->buf;
765 len = dst->len;
6073ee85 766 }
3bfba20d 767 return ret | ident_to_git(path, src, len, dst, ca.ident);
35ebfd6a
JH
768}
769
43dd2332
EB
770static int convert_to_working_tree_internal(const char *path, const char *src,
771 size_t len, struct strbuf *dst,
772 int normalizing)
35ebfd6a 773{
3bfba20d 774 int ret = 0;
cd8be6c9 775 const char *filter = NULL;
3bfba20d 776 struct conv_attrs ca;
6073ee85 777
3bfba20d
JH
778 convert_attrs(&ca, path);
779 if (ca.drv)
780 filter = ca.drv->smudge;
3fed15f5 781
3bfba20d 782 ret |= ident_to_worktree(path, src, len, dst, ca.ident);
5ecd293d
PH
783 if (ret) {
784 src = dst->buf;
785 len = dst->len;
3fed15f5 786 }
43dd2332
EB
787 /*
788 * CRLF conversion can be skipped if normalizing, unless there
789 * is a smudge filter. The filter might expect CRLFs.
790 */
791 if (filter || !normalizing) {
3bfba20d
JH
792 ca.crlf_action = input_crlf_action(ca.crlf_action, ca.eol_attr);
793 ret |= crlf_to_worktree(path, src, len, dst, ca.crlf_action);
43dd2332
EB
794 if (ret) {
795 src = dst->buf;
796 len = dst->len;
797 }
aa4ed402 798 }
5ecd293d 799 return ret | apply_filter(path, src, len, dst, filter);
35ebfd6a 800}
f217f0e8 801
43dd2332
EB
802int convert_to_working_tree(const char *path, const char *src, size_t len, struct strbuf *dst)
803{
804 return convert_to_working_tree_internal(path, src, len, dst, 0);
805}
806
f217f0e8
EB
807int renormalize_buffer(const char *path, const char *src, size_t len, struct strbuf *dst)
808{
43dd2332 809 int ret = convert_to_working_tree_internal(path, src, len, dst, 1);
f217f0e8
EB
810 if (ret) {
811 src = dst->buf;
812 len = dst->len;
813 }
814 return ret | convert_to_git(path, src, len, dst, 0);
815}
dd8e9121 816
b6691092
JH
817/*****************************************************************
818 *
819 * Streaming converison support
820 *
821 *****************************************************************/
822
823typedef int (*filter_fn)(struct stream_filter *,
824 const char *input, size_t *isize_p,
825 char *output, size_t *osize_p);
826typedef void (*free_fn)(struct stream_filter *);
827
828struct stream_filter_vtbl {
829 filter_fn filter;
830 free_fn free;
831};
832
833struct stream_filter {
834 struct stream_filter_vtbl *vtbl;
835};
836
837static int null_filter_fn(struct stream_filter *filter,
838 const char *input, size_t *isize_p,
839 char *output, size_t *osize_p)
840{
4ae66704
JH
841 size_t count;
842
843 if (!input)
844 return 0; /* we do not keep any states */
845 count = *isize_p;
b6691092
JH
846 if (*osize_p < count)
847 count = *osize_p;
848 if (count) {
849 memmove(output, input, count);
850 *isize_p -= count;
851 *osize_p -= count;
852 }
853 return 0;
854}
855
856static void null_free_fn(struct stream_filter *filter)
857{
858 ; /* nothing -- null instances are shared */
859}
860
861static struct stream_filter_vtbl null_vtbl = {
862 null_filter_fn,
863 null_free_fn,
864};
865
866static struct stream_filter null_filter_singleton = {
867 &null_vtbl,
868};
869
870int is_null_stream_filter(struct stream_filter *filter)
871{
872 return filter == &null_filter_singleton;
873}
874
b84c7839
JH
875
876/*
877 * LF-to-CRLF filter
878 */
284e3d28
CMN
879
880struct lf_to_crlf_filter {
881 struct stream_filter filter;
87afe9a5 882 unsigned want_lf:1;
284e3d28
CMN
883};
884
e322ee38
JH
885static int lf_to_crlf_filter_fn(struct stream_filter *filter,
886 const char *input, size_t *isize_p,
887 char *output, size_t *osize_p)
888{
284e3d28
CMN
889 size_t count, o = 0;
890 struct lf_to_crlf_filter *lf_to_crlf = (struct lf_to_crlf_filter *)filter;
891
892 /* Output a pending LF if we need to */
893 if (lf_to_crlf->want_lf) {
894 output[o++] = '\n';
895 lf_to_crlf->want_lf = 0;
896 }
e322ee38 897
87afe9a5
JH
898 /* We are told to drain */
899 if (!input) {
900 *osize_p -= o;
901 return 0;
902 }
284e3d28 903
e322ee38
JH
904 count = *isize_p;
905 if (count) {
284e3d28
CMN
906 size_t i;
907 for (i = 0; o < *osize_p && i < count; i++) {
e322ee38
JH
908 char ch = input[i];
909 if (ch == '\n') {
284e3d28
CMN
910 output[o++] = '\r';
911 if (o >= *osize_p) {
912 lf_to_crlf->want_lf = 1;
913 continue; /* We need to increase i */
914 }
e322ee38
JH
915 }
916 output[o++] = ch;
917 }
918
919 *osize_p -= o;
920 *isize_p -= i;
921 }
922 return 0;
923}
924
284e3d28
CMN
925static void lf_to_crlf_free_fn(struct stream_filter *filter)
926{
927 free(filter);
928}
929
e322ee38
JH
930static struct stream_filter_vtbl lf_to_crlf_vtbl = {
931 lf_to_crlf_filter_fn,
284e3d28 932 lf_to_crlf_free_fn,
e322ee38
JH
933};
934
284e3d28
CMN
935static struct stream_filter *lf_to_crlf_filter(void)
936{
87afe9a5 937 struct lf_to_crlf_filter *lf_to_crlf = xcalloc(1, sizeof(*lf_to_crlf));
e322ee38 938
284e3d28 939 lf_to_crlf->filter.vtbl = &lf_to_crlf_vtbl;
284e3d28
CMN
940 return (struct stream_filter *)lf_to_crlf;
941}
b84c7839 942
a265a7f9
JH
943/*
944 * Cascade filter
945 */
946#define FILTER_BUFFER 1024
947struct cascade_filter {
948 struct stream_filter filter;
949 struct stream_filter *one;
950 struct stream_filter *two;
951 char buf[FILTER_BUFFER];
952 int end, ptr;
953};
954
955static int cascade_filter_fn(struct stream_filter *filter,
956 const char *input, size_t *isize_p,
957 char *output, size_t *osize_p)
958{
959 struct cascade_filter *cas = (struct cascade_filter *) filter;
960 size_t filled = 0;
961 size_t sz = *osize_p;
962 size_t to_feed, remaining;
963
964 /*
965 * input -- (one) --> buf -- (two) --> output
966 */
967 while (filled < sz) {
968 remaining = sz - filled;
969
970 /* do we already have something to feed two with? */
971 if (cas->ptr < cas->end) {
972 to_feed = cas->end - cas->ptr;
973 if (stream_filter(cas->two,
974 cas->buf + cas->ptr, &to_feed,
975 output + filled, &remaining))
976 return -1;
977 cas->ptr += (cas->end - cas->ptr) - to_feed;
978 filled = sz - remaining;
979 continue;
980 }
981
982 /* feed one from upstream and have it emit into our buffer */
983 to_feed = input ? *isize_p : 0;
984 if (input && !to_feed)
985 break;
986 remaining = sizeof(cas->buf);
987 if (stream_filter(cas->one,
988 input, &to_feed,
989 cas->buf, &remaining))
990 return -1;
991 cas->end = sizeof(cas->buf) - remaining;
992 cas->ptr = 0;
993 if (input) {
994 size_t fed = *isize_p - to_feed;
995 *isize_p -= fed;
996 input += fed;
997 }
998
999 /* do we know that we drained one completely? */
1000 if (input || cas->end)
1001 continue;
1002
1003 /* tell two to drain; we have nothing more to give it */
1004 to_feed = 0;
1005 remaining = sz - filled;
1006 if (stream_filter(cas->two,
1007 NULL, &to_feed,
1008 output + filled, &remaining))
1009 return -1;
1010 if (remaining == (sz - filled))
1011 break; /* completely drained two */
1012 filled = sz - remaining;
1013 }
1014 *osize_p -= filled;
1015 return 0;
1016}
1017
1018static void cascade_free_fn(struct stream_filter *filter)
1019{
1020 struct cascade_filter *cas = (struct cascade_filter *)filter;
1021 free_stream_filter(cas->one);
1022 free_stream_filter(cas->two);
1023 free(filter);
1024}
1025
1026static struct stream_filter_vtbl cascade_vtbl = {
1027 cascade_filter_fn,
1028 cascade_free_fn,
1029};
1030
1031static struct stream_filter *cascade_filter(struct stream_filter *one,
1032 struct stream_filter *two)
1033{
1034 struct cascade_filter *cascade;
1035
1036 if (!one || is_null_stream_filter(one))
1037 return two;
1038 if (!two || is_null_stream_filter(two))
1039 return one;
1040
1041 cascade = xmalloc(sizeof(*cascade));
1042 cascade->one = one;
1043 cascade->two = two;
1044 cascade->end = cascade->ptr = 0;
1045 cascade->filter.vtbl = &cascade_vtbl;
1046 return (struct stream_filter *)cascade;
1047}
1048
b84c7839
JH
1049/*
1050 * ident filter
1051 */
1052#define IDENT_DRAINING (-1)
1053#define IDENT_SKIPPING (-2)
1054struct ident_filter {
1055 struct stream_filter filter;
1056 struct strbuf left;
1057 int state;
1058 char ident[45]; /* ": x40 $" */
1059};
1060
1061static int is_foreign_ident(const char *str)
1062{
1063 int i;
1064
1065 if (prefixcmp(str, "$Id: "))
1066 return 0;
1067 for (i = 5; str[i]; i++) {
1068 if (isspace(str[i]) && str[i+1] != '$')
1069 return 1;
1070 }
1071 return 0;
1072}
1073
1074static void ident_drain(struct ident_filter *ident, char **output_p, size_t *osize_p)
1075{
1076 size_t to_drain = ident->left.len;
1077
1078 if (*osize_p < to_drain)
1079 to_drain = *osize_p;
1080 if (to_drain) {
1081 memcpy(*output_p, ident->left.buf, to_drain);
1082 strbuf_remove(&ident->left, 0, to_drain);
1083 *output_p += to_drain;
1084 *osize_p -= to_drain;
1085 }
1086 if (!ident->left.len)
1087 ident->state = 0;
1088}
1089
1090static int ident_filter_fn(struct stream_filter *filter,
1091 const char *input, size_t *isize_p,
1092 char *output, size_t *osize_p)
1093{
1094 struct ident_filter *ident = (struct ident_filter *)filter;
1095 static const char head[] = "$Id";
1096
1097 if (!input) {
1098 /* drain upon eof */
1099 switch (ident->state) {
1100 default:
1101 strbuf_add(&ident->left, head, ident->state);
1102 case IDENT_SKIPPING:
1103 /* fallthru */
1104 case IDENT_DRAINING:
1105 ident_drain(ident, &output, osize_p);
1106 }
1107 return 0;
1108 }
1109
1110 while (*isize_p || (ident->state == IDENT_DRAINING)) {
1111 int ch;
1112
1113 if (ident->state == IDENT_DRAINING) {
1114 ident_drain(ident, &output, osize_p);
1115 if (!*osize_p)
1116 break;
1117 continue;
1118 }
1119
1120 ch = *(input++);
1121 (*isize_p)--;
1122
1123 if (ident->state == IDENT_SKIPPING) {
1124 /*
1125 * Skipping until '$' or LF, but keeping them
1126 * in case it is a foreign ident.
1127 */
1128 strbuf_addch(&ident->left, ch);
1129 if (ch != '\n' && ch != '$')
1130 continue;
1131 if (ch == '$' && !is_foreign_ident(ident->left.buf)) {
1132 strbuf_setlen(&ident->left, sizeof(head) - 1);
1133 strbuf_addstr(&ident->left, ident->ident);
1134 }
1135 ident->state = IDENT_DRAINING;
1136 continue;
1137 }
1138
1139 if (ident->state < sizeof(head) &&
1140 head[ident->state] == ch) {
1141 ident->state++;
1142 continue;
1143 }
1144
1145 if (ident->state)
1146 strbuf_add(&ident->left, head, ident->state);
1147 if (ident->state == sizeof(head) - 1) {
1148 if (ch != ':' && ch != '$') {
1149 strbuf_addch(&ident->left, ch);
1150 ident->state = 0;
1151 continue;
1152 }
1153
1154 if (ch == ':') {
1155 strbuf_addch(&ident->left, ch);
1156 ident->state = IDENT_SKIPPING;
1157 } else {
1158 strbuf_addstr(&ident->left, ident->ident);
1159 ident->state = IDENT_DRAINING;
1160 }
1161 continue;
1162 }
1163
1164 strbuf_addch(&ident->left, ch);
1165 ident->state = IDENT_DRAINING;
1166 }
1167 return 0;
1168}
1169
1170static void ident_free_fn(struct stream_filter *filter)
1171{
1172 struct ident_filter *ident = (struct ident_filter *)filter;
1173 strbuf_release(&ident->left);
1174 free(filter);
1175}
1176
1177static struct stream_filter_vtbl ident_vtbl = {
1178 ident_filter_fn,
1179 ident_free_fn,
1180};
1181
1182static struct stream_filter *ident_filter(const unsigned char *sha1)
1183{
1184 struct ident_filter *ident = xmalloc(sizeof(*ident));
1185
1186 sprintf(ident->ident, ": %s $", sha1_to_hex(sha1));
1187 strbuf_init(&ident->left, 0);
1188 ident->filter.vtbl = &ident_vtbl;
1189 ident->state = 0;
1190 return (struct stream_filter *)ident;
1191}
1192
dd8e9121 1193/*
b6691092
JH
1194 * Return an appropriately constructed filter for the path, or NULL if
1195 * the contents cannot be filtered without reading the whole thing
1196 * in-core.
1197 *
1198 * Note that you would be crazy to set CRLF, smuge/clean or ident to a
1199 * large binary blob you would want us not to slurp into the memory!
dd8e9121 1200 */
b6691092 1201struct stream_filter *get_stream_filter(const char *path, const unsigned char *sha1)
dd8e9121
JH
1202{
1203 struct conv_attrs ca;
1204 enum crlf_action crlf_action;
b84c7839 1205 struct stream_filter *filter = NULL;
dd8e9121
JH
1206
1207 convert_attrs(&ca, path);
1208
b84c7839
JH
1209 if (ca.drv && (ca.drv->smudge || ca.drv->clean))
1210 return filter;
1211
1212 if (ca.ident)
1213 filter = ident_filter(sha1);
dd8e9121
JH
1214
1215 crlf_action = input_crlf_action(ca.crlf_action, ca.eol_attr);
b84c7839 1216
b0d9c69f 1217 if ((crlf_action == CRLF_BINARY) || (crlf_action == CRLF_INPUT) ||
a265a7f9
JH
1218 (crlf_action == CRLF_GUESS && auto_crlf == AUTO_CRLF_FALSE))
1219 filter = cascade_filter(filter, &null_filter_singleton);
1220
1221 else if (output_eol(crlf_action) == EOL_CRLF &&
1222 !(crlf_action == CRLF_AUTO || crlf_action == CRLF_GUESS))
284e3d28 1223 filter = cascade_filter(filter, lf_to_crlf_filter());
e322ee38 1224
b84c7839 1225 return filter;
b6691092
JH
1226}
1227
1228void free_stream_filter(struct stream_filter *filter)
1229{
1230 filter->vtbl->free(filter);
1231}
1232
1233int stream_filter(struct stream_filter *filter,
1234 const char *input, size_t *isize_p,
1235 char *output, size_t *osize_p)
1236{
1237 return filter->vtbl->filter(filter, input, isize_p, output, osize_p);
dd8e9121 1238}