convert: make apply_filter() adhere to standard Git error handling
[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"
6424c2ad 5#include "sigchain.h"
35ebfd6a 6
6c510bee
LT
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
942e7747 13 * translation when the "text" attribute or "auto_crlf" option is set.
6c510bee
LT
14 */
15
a7630bd4
TB
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
c61dcff9 21enum crlf_action {
df747b81
TB
22 CRLF_UNDEFINED,
23 CRLF_BINARY,
fd6cce9e 24 CRLF_TEXT,
df747b81
TB
25 CRLF_TEXT_INPUT,
26 CRLF_TEXT_CRLF,
27 CRLF_AUTO,
28 CRLF_AUTO_INPUT,
29 CRLF_AUTO_CRLF
fd6cce9e 30};
163b9591 31
6c510bee 32struct text_stat {
28624193 33 /* NUL, CR, LF and CRLF counts */
6e336a53 34 unsigned nul, lonecr, lonelf, crlf;
6c510bee
LT
35
36 /* These are just approximations! */
37 unsigned printable, nonprintable;
38};
39
40static 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') {
6e336a53 49 if (i+1 < size && buf[i+1] == '\n') {
6c510bee 50 stats->crlf++;
6e336a53
TB
51 i++;
52 } else
53 stats->lonecr++;
6c510bee
LT
54 continue;
55 }
56 if (c == '\n') {
6e336a53 57 stats->lonelf++;
6c510bee
LT
58 continue;
59 }
60 if (c == 127)
61 /* DEL */
62 stats->nonprintable++;
63 else if (c < 32) {
64 switch (c) {
65 /* BS, HT, ESC and FF */
66 case '\b': case '\t': case '\033': case '\014':
67 stats->printable++;
68 break;
28624193
DP
69 case 0:
70 stats->nul++;
71 /* fall through */
6c510bee
LT
72 default:
73 stats->nonprintable++;
74 }
75 }
76 else
77 stats->printable++;
78 }
f9dd4bf4
DK
79
80 /* If file ends with EOF then don't count this EOF as non-printable. */
81 if (size >= 1 && buf[size-1] == '\032')
82 stats->nonprintable--;
6c510bee
LT
83}
84
85/*
86 * The same heuristics as diff.c::mmfile_is_binary()
a7630bd4 87 * We treat files with bare CR as binary
6c510bee 88 */
a7630bd4 89static int convert_is_binary(unsigned long size, const struct text_stat *stats)
6c510bee 90{
6e336a53 91 if (stats->lonecr)
a7630bd4 92 return 1;
28624193
DP
93 if (stats->nul)
94 return 1;
6c510bee
LT
95 if ((stats->printable >> 7) < stats->nonprintable)
96 return 1;
6c510bee
LT
97 return 0;
98}
99
a7630bd4
TB
100static unsigned int gather_convert_stats(const char *data, unsigned long size)
101{
102 struct text_stat stats;
6e336a53 103 int ret = 0;
a7630bd4
TB
104 if (!data || !size)
105 return 0;
106 gather_stats(data, size, &stats);
107 if (convert_is_binary(size, &stats))
6e336a53
TB
108 ret |= CONVERT_STAT_BITS_BIN;
109 if (stats.crlf)
110 ret |= CONVERT_STAT_BITS_TXT_CRLF;
111 if (stats.lonelf)
112 ret |= CONVERT_STAT_BITS_TXT_LF;
113
114 return ret;
a7630bd4
TB
115}
116
117static const char *gather_convert_stats_ascii(const char *data, unsigned long size)
118{
119 unsigned int convert_stats = gather_convert_stats(data, size);
120
121 if (convert_stats & CONVERT_STAT_BITS_BIN)
122 return "-text";
123 switch (convert_stats) {
124 case CONVERT_STAT_BITS_TXT_LF:
125 return "lf";
126 case CONVERT_STAT_BITS_TXT_CRLF:
127 return "crlf";
128 case CONVERT_STAT_BITS_TXT_LF | CONVERT_STAT_BITS_TXT_CRLF:
129 return "mixed";
130 default:
131 return "none";
132 }
133}
134
135const char *get_cached_convert_stats_ascii(const char *path)
136{
137 const char *ret;
138 unsigned long sz;
139 void *data = read_blob_data_from_cache(path, &sz);
140 ret = gather_convert_stats_ascii(data, sz);
141 free(data);
142 return ret;
143}
144
145const char *get_wt_convert_stats_ascii(const char *path)
146{
147 const char *ret = "";
148 struct strbuf sb = STRBUF_INIT;
149 if (strbuf_read_file(&sb, path, 0) >= 0)
150 ret = gather_convert_stats_ascii(sb.buf, sb.len);
151 strbuf_release(&sb);
152 return ret;
153}
154
4b4024f5
TB
155static int text_eol_is_crlf(void)
156{
157 if (auto_crlf == AUTO_CRLF_TRUE)
158 return 1;
159 else if (auto_crlf == AUTO_CRLF_INPUT)
160 return 0;
161 if (core_eol == EOL_CRLF)
162 return 1;
163 if (core_eol == EOL_UNSET && EOL_NATIVE == EOL_CRLF)
164 return 1;
165 return 0;
166}
167
c61dcff9 168static enum eol output_eol(enum crlf_action crlf_action)
f217f0e8 169{
c61dcff9 170 switch (crlf_action) {
942e7747
EB
171 case CRLF_BINARY:
172 return EOL_UNSET;
df747b81 173 case CRLF_TEXT_CRLF:
942e7747 174 return EOL_CRLF;
df747b81 175 case CRLF_TEXT_INPUT:
942e7747 176 return EOL_LF;
df747b81
TB
177 case CRLF_UNDEFINED:
178 case CRLF_AUTO_CRLF:
65237284 179 return EOL_CRLF;
df747b81 180 case CRLF_AUTO_INPUT:
65237284 181 return EOL_LF;
942e7747
EB
182 case CRLF_TEXT:
183 case CRLF_AUTO:
df747b81 184 /* fall through */
4b4024f5 185 return text_eol_is_crlf() ? EOL_CRLF : EOL_LF;
942e7747 186 }
df747b81 187 warning("Illegal crlf_action %d\n", (int)crlf_action);
ec70f52f 188 return core_eol;
942e7747
EB
189}
190
c61dcff9 191static void check_safe_crlf(const char *path, enum crlf_action crlf_action,
a0ad53c1
TB
192 struct text_stat *old_stats, struct text_stat *new_stats,
193 enum safe_crlf checksafe)
21e5ad50 194{
a0ad53c1 195 if (old_stats->crlf && !new_stats->crlf ) {
21e5ad50 196 /*
a0ad53c1 197 * CRLFs would not be restored by checkout
21e5ad50 198 */
a0ad53c1
TB
199 if (checksafe == SAFE_CRLF_WARN)
200 warning("CRLF will be replaced by LF in %s.\nThe file will have its original line endings in your working directory.", path);
201 else /* i.e. SAFE_CRLF_FAIL */
202 die("CRLF would be replaced by LF in %s.", path);
203 } else if (old_stats->lonelf && !new_stats->lonelf ) {
21e5ad50 204 /*
a0ad53c1 205 * CRLFs would be added by checkout
21e5ad50 206 */
a0ad53c1
TB
207 if (checksafe == SAFE_CRLF_WARN)
208 warning("LF will be replaced by CRLF in %s.\nThe file will have its original line endings in your working directory.", path);
209 else /* i.e. SAFE_CRLF_FAIL */
210 die("LF would be replaced by CRLF in %s", path);
21e5ad50
SP
211 }
212}
213
c4805393
FAG
214static int has_cr_in_index(const char *path)
215{
c4805393 216 unsigned long sz;
c4805393
FAG
217 void *data;
218 int has_cr;
c4805393 219
4982fd78
LF
220 data = read_blob_data_from_cache(path, &sz);
221 if (!data)
c4805393 222 return 0;
c4805393
FAG
223 has_cr = memchr(data, '\r', sz) != NULL;
224 free(data);
225 return has_cr;
226}
227
a0ad53c1
TB
228static int will_convert_lf_to_crlf(size_t len, struct text_stat *stats,
229 enum crlf_action crlf_action)
230{
231 if (output_eol(crlf_action) != EOL_CRLF)
232 return 0;
233 /* No "naked" LF? Nothing to convert, regardless. */
234 if (!stats->lonelf)
235 return 0;
236
237 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
238 /* If we have any CR or CRLF line endings, we do not touch it */
239 /* This is the new safer autocrlf-handling */
240 if (stats->lonecr || stats->crlf)
241 return 0;
242
243 if (convert_is_binary(len, stats))
244 return 0;
245 }
246 return 1;
247
248}
249
5ecd293d 250static int crlf_to_git(const char *path, const char *src, size_t len,
c61dcff9
JH
251 struct strbuf *buf,
252 enum crlf_action crlf_action, enum safe_crlf checksafe)
6c510bee 253{
6c510bee 254 struct text_stat stats;
5ecd293d 255 char *dst;
a0ad53c1 256 int convert_crlf_into_lf;
6c510bee 257
c61dcff9 258 if (crlf_action == CRLF_BINARY ||
4c3b57b9 259 (src && !len))
5ecd293d 260 return 0;
6c510bee 261
4c3b57b9
JK
262 /*
263 * If we are doing a dry-run and have no source buffer, there is
264 * nothing to analyze; we must assume we would convert.
265 */
266 if (!buf && !src)
267 return 1;
268
5ecd293d 269 gather_stats(src, len, &stats);
a0ad53c1
TB
270 /* Optimization: No CRLF? Nothing to convert, regardless. */
271 convert_crlf_into_lf = !!stats.crlf;
6c510bee 272
df747b81 273 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
a7630bd4 274 if (convert_is_binary(len, &stats))
5ecd293d 275 return 0;
65237284
TB
276 /*
277 * If the file in the index has any CR in it, do not convert.
278 * This is the new safer autocrlf handling.
279 */
280 if (checksafe == SAFE_CRLF_RENORMALIZE)
281 checksafe = SAFE_CRLF_FALSE;
282 else if (has_cr_in_index(path))
a0ad53c1 283 convert_crlf_into_lf = 0;
201ac8ef 284 }
a0ad53c1
TB
285 if (checksafe && len) {
286 struct text_stat new_stats;
287 memcpy(&new_stats, &stats, sizeof(new_stats));
288 /* simulate "git add" */
289 if (convert_crlf_into_lf) {
290 new_stats.lonelf += new_stats.crlf;
291 new_stats.crlf = 0;
292 }
293 /* simulate "git checkout" */
294 if (will_convert_lf_to_crlf(len, &new_stats, crlf_action)) {
295 new_stats.crlf += new_stats.lonelf;
296 new_stats.lonelf = 0;
297 }
298 check_safe_crlf(path, crlf_action, &stats, &new_stats, checksafe);
299 }
300 if (!convert_crlf_into_lf)
21e5ad50
SP
301 return 0;
302
92ac3197
JK
303 /*
304 * At this point all of our source analysis is done, and we are sure we
305 * would convert. If we are in dry-run mode, we can give an answer.
306 */
307 if (!buf)
308 return 1;
309
90d16ec0
PH
310 /* only grow if not in place */
311 if (strbuf_avail(buf) + buf->len < len)
312 strbuf_grow(buf, len - buf->len);
5ecd293d 313 dst = buf->buf;
df747b81 314 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
163b9591
JH
315 /*
316 * If we guessed, we already know we rejected a file with
317 * lone CR, and we can strip a CR without looking at what
318 * follow it.
319 */
201ac8ef 320 do {
ac78e548 321 unsigned char c = *src++;
201ac8ef 322 if (c != '\r')
ac78e548 323 *dst++ = c;
5ecd293d 324 } while (--len);
201ac8ef
JH
325 } else {
326 do {
ac78e548 327 unsigned char c = *src++;
5ecd293d 328 if (! (c == '\r' && (1 < len && *src == '\n')))
ac78e548 329 *dst++ = c;
5ecd293d 330 } while (--len);
201ac8ef 331 }
5ecd293d
PH
332 strbuf_setlen(buf, dst - buf->buf);
333 return 1;
6c510bee
LT
334}
335
5ecd293d 336static int crlf_to_worktree(const char *path, const char *src, size_t len,
c61dcff9 337 struct strbuf *buf, enum crlf_action crlf_action)
6c510bee 338{
5ecd293d 339 char *to_free = NULL;
6c510bee 340 struct text_stat stats;
6c510bee 341
c61dcff9 342 if (!len || output_eol(crlf_action) != EOL_CRLF)
5ecd293d 343 return 0;
6c510bee 344
5ecd293d 345 gather_stats(src, len, &stats);
a0ad53c1 346 if (!will_convert_lf_to_crlf(len, &stats, crlf_action))
5ecd293d 347 return 0;
6c510bee 348
5ecd293d
PH
349 /* are we "faking" in place editing ? */
350 if (src == buf->buf)
b315c5c0 351 to_free = strbuf_detach(buf, NULL);
5ecd293d 352
6e336a53 353 strbuf_grow(buf, len + stats.lonelf);
5ecd293d
PH
354 for (;;) {
355 const char *nl = memchr(src, '\n', len);
356 if (!nl)
357 break;
358 if (nl > src && nl[-1] == '\r') {
359 strbuf_add(buf, src, nl + 1 - src);
360 } else {
361 strbuf_add(buf, src, nl - src);
362 strbuf_addstr(buf, "\r\n");
363 }
364 len -= nl + 1 - src;
365 src = nl + 1;
366 }
367 strbuf_add(buf, src, len);
368
369 free(to_free);
370 return 1;
6c510bee 371}
35ebfd6a 372
546bb582
JS
373struct filter_params {
374 const char *src;
375 unsigned long size;
9035d75a 376 int fd;
546bb582 377 const char *cmd;
a2b665de 378 const char *path;
546bb582
JS
379};
380
9035d75a 381static int filter_buffer_or_fd(int in, int out, void *data)
aa4ed402
JH
382{
383 /*
384 * Spawn cmd and feed the buffer contents through its stdin.
385 */
d3180279 386 struct child_process child_process = CHILD_PROCESS_INIT;
546bb582 387 struct filter_params *params = (struct filter_params *)data;
aa4ed402 388 int write_err, status;
66dbfd55
GV
389 const char *argv[] = { NULL, NULL };
390
a2b665de
PW
391 /* apply % substitution to cmd */
392 struct strbuf cmd = STRBUF_INIT;
393 struct strbuf path = STRBUF_INIT;
394 struct strbuf_expand_dict_entry dict[] = {
395 { "f", NULL, },
396 { NULL, NULL, },
397 };
398
399 /* quote the path to preserve spaces, etc. */
400 sq_quote_buf(&path, params->path);
401 dict[0].value = path.buf;
402
403 /* expand all %f with the quoted path */
404 strbuf_expand(&cmd, params->cmd, strbuf_expand_dict_cb, &dict);
405 strbuf_release(&path);
406
407 argv[0] = cmd.buf;
aa4ed402 408
dc1bfdcd 409 child_process.argv = argv;
ac0ba18d 410 child_process.use_shell = 1;
dc1bfdcd 411 child_process.in = -1;
ae6a5609 412 child_process.out = out;
aa4ed402 413
dc1bfdcd 414 if (start_command(&child_process))
255f04d6 415 return error("cannot fork to run external filter '%s'", params->cmd);
aa4ed402 416
6424c2ad
JB
417 sigchain_push(SIGPIPE, SIG_IGN);
418
9035d75a 419 if (params->src) {
0c4dd67a
JH
420 write_err = (write_in_full(child_process.in,
421 params->src, params->size) < 0);
422 if (errno == EPIPE)
423 write_err = 0;
9035d75a
SP
424 } else {
425 write_err = copy_fd(params->fd, child_process.in);
0c4dd67a
JH
426 if (write_err == COPY_WRITE_ERROR && errno == EPIPE)
427 write_err = 0;
9035d75a
SP
428 }
429
dc1bfdcd 430 if (close(child_process.in))
aa4ed402
JH
431 write_err = 1;
432 if (write_err)
255f04d6 433 error("cannot feed the input to external filter '%s'", params->cmd);
aa4ed402 434
6424c2ad
JB
435 sigchain_pop(SIGPIPE);
436
aa4ed402
JH
437 status = finish_command(&child_process);
438 if (status)
255f04d6 439 error("external filter '%s' failed %d", params->cmd, status);
a2b665de
PW
440
441 strbuf_release(&cmd);
aa4ed402
JH
442 return (write_err || status);
443}
444
9035d75a 445static int apply_filter(const char *path, const char *src, size_t len, int fd,
5ecd293d 446 struct strbuf *dst, const char *cmd)
aa4ed402
JH
447{
448 /*
449 * Create a pipeline to have the command filter the buffer's
450 * contents.
451 *
452 * (child --> cmd) --> us
453 */
b84be553 454 int err = 0;
f285a2d7 455 struct strbuf nbuf = STRBUF_INIT;
546bb582
JS
456 struct async async;
457 struct filter_params params;
aa4ed402 458
1a8630dc 459 if (!cmd || !*cmd)
5ecd293d 460 return 0;
aa4ed402 461
92ac3197
JK
462 if (!dst)
463 return 1;
464
546bb582 465 memset(&async, 0, sizeof(async));
9035d75a 466 async.proc = filter_buffer_or_fd;
546bb582 467 async.data = &params;
ae6a5609 468 async.out = -1;
546bb582
JS
469 params.src = src;
470 params.size = len;
9035d75a 471 params.fd = fd;
546bb582 472 params.cmd = cmd;
a2b665de 473 params.path = path;
aa4ed402
JH
474
475 fflush(NULL);
546bb582
JS
476 if (start_async(&async))
477 return 0; /* error was already reported */
aa4ed402 478
546bb582 479 if (strbuf_read(&nbuf, async.out, len) < 0) {
b84be553 480 err = error("read from external filter '%s' failed", cmd);
aa4ed402 481 }
546bb582 482 if (close(async.out)) {
b84be553 483 err = error("read from external filter '%s' failed", cmd);
aa4ed402 484 }
546bb582 485 if (finish_async(&async)) {
b84be553 486 err = error("external filter '%s' failed", cmd);
aa4ed402
JH
487 }
488
b84be553 489 if (!err) {
90d16ec0 490 strbuf_swap(dst, &nbuf);
5ecd293d 491 }
90d16ec0 492 strbuf_release(&nbuf);
b84be553 493 return !err;
aa4ed402
JH
494}
495
496static struct convert_driver {
497 const char *name;
498 struct convert_driver *next;
cd8be6c9
BH
499 const char *smudge;
500 const char *clean;
36daaaca 501 int required;
aa4ed402
JH
502} *user_convert, **user_convert_tail;
503
ef90d6d4 504static int read_convert_config(const char *var, const char *value, void *cb)
aa4ed402 505{
d731f0ad 506 const char *key, *name;
aa4ed402
JH
507 int namelen;
508 struct convert_driver *drv;
509
510 /*
511 * External conversion drivers are configured using
512 * "filter.<name>.variable".
513 */
d731f0ad 514 if (parse_config_key(var, "filter", &name, &namelen, &key) < 0 || !name)
aa4ed402 515 return 0;
aa4ed402
JH
516 for (drv = user_convert; drv; drv = drv->next)
517 if (!strncmp(drv->name, name, namelen) && !drv->name[namelen])
518 break;
519 if (!drv) {
aa4ed402 520 drv = xcalloc(1, sizeof(struct convert_driver));
182af834 521 drv->name = xmemdupz(name, namelen);
aa4ed402
JH
522 *user_convert_tail = drv;
523 user_convert_tail = &(drv->next);
524 }
525
aa4ed402
JH
526 /*
527 * filter.<name>.smudge and filter.<name>.clean specifies
528 * the command line:
529 *
530 * command-line
531 *
532 * The command-line will not be interpolated in any way.
533 */
534
d731f0ad 535 if (!strcmp("smudge", key))
cd8be6c9
BH
536 return git_config_string(&drv->smudge, var, value);
537
d731f0ad 538 if (!strcmp("clean", key))
cd8be6c9 539 return git_config_string(&drv->clean, var, value);
aa4ed402 540
d731f0ad 541 if (!strcmp("required", key)) {
36daaaca
JB
542 drv->required = git_config_bool(var, value);
543 return 0;
544 }
545
aa4ed402
JH
546 return 0;
547}
548
3fed15f5
JH
549static int count_ident(const char *cp, unsigned long size)
550{
551 /*
af9b54bb 552 * "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$"
3fed15f5
JH
553 */
554 int cnt = 0;
555 char ch;
556
557 while (size) {
558 ch = *cp++;
559 size--;
560 if (ch != '$')
561 continue;
af9b54bb 562 if (size < 3)
3fed15f5 563 break;
af9b54bb 564 if (memcmp("Id", cp, 2))
3fed15f5 565 continue;
af9b54bb
AP
566 ch = cp[2];
567 cp += 3;
568 size -= 3;
3fed15f5 569 if (ch == '$')
af9b54bb 570 cnt++; /* $Id$ */
3fed15f5
JH
571 if (ch != ':')
572 continue;
573
574 /*
af9b54bb 575 * "$Id: ... "; scan up to the closing dollar sign and discard.
3fed15f5
JH
576 */
577 while (size) {
578 ch = *cp++;
579 size--;
580 if (ch == '$') {
581 cnt++;
582 break;
583 }
a9f3049f
HG
584 if (ch == '\n')
585 break;
3fed15f5
JH
586 }
587 }
588 return cnt;
589}
590
5ecd293d
PH
591static int ident_to_git(const char *path, const char *src, size_t len,
592 struct strbuf *buf, int ident)
3fed15f5 593{
5ecd293d 594 char *dst, *dollar;
3fed15f5 595
4c3b57b9 596 if (!ident || (src && !count_ident(src, len)))
5ecd293d
PH
597 return 0;
598
92ac3197
JK
599 if (!buf)
600 return 1;
601
90d16ec0
PH
602 /* only grow if not in place */
603 if (strbuf_avail(buf) + buf->len < len)
604 strbuf_grow(buf, len - buf->len);
5ecd293d
PH
605 dst = buf->buf;
606 for (;;) {
607 dollar = memchr(src, '$', len);
608 if (!dollar)
609 break;
77321184 610 memmove(dst, src, dollar + 1 - src);
5ecd293d
PH
611 dst += dollar + 1 - src;
612 len -= dollar + 1 - src;
613 src = dollar + 1;
614
615 if (len > 3 && !memcmp(src, "Id:", 3)) {
616 dollar = memchr(src + 3, '$', len - 3);
617 if (!dollar)
618 break;
a9f3049f
HG
619 if (memchr(src + 3, '\n', dollar - src - 3)) {
620 /* Line break before the next dollar. */
621 continue;
622 }
623
af9b54bb
AP
624 memcpy(dst, "Id$", 3);
625 dst += 3;
5ecd293d
PH
626 len -= dollar + 1 - src;
627 src = dollar + 1;
3fed15f5
JH
628 }
629 }
77321184 630 memmove(dst, src, len);
5ecd293d
PH
631 strbuf_setlen(buf, dst + len - buf->buf);
632 return 1;
3fed15f5
JH
633}
634
5ecd293d
PH
635static int ident_to_worktree(const char *path, const char *src, size_t len,
636 struct strbuf *buf, int ident)
3fed15f5 637{
3fed15f5 638 unsigned char sha1[20];
07814d90 639 char *to_free = NULL, *dollar, *spc;
5ecd293d 640 int cnt;
3fed15f5
JH
641
642 if (!ident)
5ecd293d 643 return 0;
3fed15f5 644
5ecd293d 645 cnt = count_ident(src, len);
3fed15f5 646 if (!cnt)
5ecd293d 647 return 0;
3fed15f5 648
5ecd293d
PH
649 /* are we "faking" in place editing ? */
650 if (src == buf->buf)
b315c5c0 651 to_free = strbuf_detach(buf, NULL);
5ecd293d 652 hash_sha1_file(src, len, "blob", sha1);
3fed15f5 653
5ecd293d
PH
654 strbuf_grow(buf, len + cnt * 43);
655 for (;;) {
656 /* step 1: run to the next '$' */
657 dollar = memchr(src, '$', len);
658 if (!dollar)
659 break;
660 strbuf_add(buf, src, dollar + 1 - src);
661 len -= dollar + 1 - src;
662 src = dollar + 1;
c23290d5 663
5ecd293d
PH
664 /* step 2: does it looks like a bit like Id:xxx$ or Id$ ? */
665 if (len < 3 || memcmp("Id", src, 2))
3fed15f5
JH
666 continue;
667
5ecd293d
PH
668 /* step 3: skip over Id$ or Id:xxxxx$ */
669 if (src[2] == '$') {
670 src += 3;
671 len -= 3;
672 } else if (src[2] == ':') {
673 /*
674 * It's possible that an expanded Id has crept its way into the
07814d90
HG
675 * repository, we cope with that by stripping the expansion out.
676 * This is probably not a good idea, since it will cause changes
677 * on checkout, which won't go away by stash, but let's keep it
678 * for git-style ids.
5ecd293d
PH
679 */
680 dollar = memchr(src + 3, '$', len - 3);
681 if (!dollar) {
682 /* incomplete keyword, no more '$', so just quit the loop */
683 break;
684 }
c23290d5 685
a9f3049f
HG
686 if (memchr(src + 3, '\n', dollar - src - 3)) {
687 /* Line break before the next dollar. */
688 continue;
689 }
690
07814d90
HG
691 spc = memchr(src + 4, ' ', dollar - src - 4);
692 if (spc && spc < dollar-1) {
693 /* There are spaces in unexpected places.
694 * This is probably an id from some other
695 * versioning system. Keep it for now.
696 */
697 continue;
698 }
699
5ecd293d
PH
700 len -= dollar + 1 - src;
701 src = dollar + 1;
702 } else {
703 /* it wasn't a "Id$" or "Id:xxxx$" */
704 continue;
705 }
c23290d5 706
5ecd293d
PH
707 /* step 4: substitute */
708 strbuf_addstr(buf, "Id: ");
709 strbuf_add(buf, sha1_to_hex(sha1), 40);
710 strbuf_addstr(buf, " $");
3fed15f5 711 }
5ecd293d 712 strbuf_add(buf, src, len);
3fed15f5 713
5ecd293d
PH
714 free(to_free);
715 return 1;
35ebfd6a
JH
716}
717
92cce135 718static enum crlf_action git_path_check_crlf(struct git_attr_check *check)
35ebfd6a 719{
6073ee85
JH
720 const char *value = check->value;
721
722 if (ATTR_TRUE(value))
723 return CRLF_TEXT;
724 else if (ATTR_FALSE(value))
725 return CRLF_BINARY;
726 else if (ATTR_UNSET(value))
727 ;
728 else if (!strcmp(value, "input"))
df747b81 729 return CRLF_TEXT_INPUT;
fd6cce9e
EB
730 else if (!strcmp(value, "auto"))
731 return CRLF_AUTO;
df747b81 732 return CRLF_UNDEFINED;
35ebfd6a
JH
733}
734
92cce135 735static enum eol git_path_check_eol(struct git_attr_check *check)
fd6cce9e
EB
736{
737 const char *value = check->value;
738
739 if (ATTR_UNSET(value))
740 ;
741 else if (!strcmp(value, "lf"))
742 return EOL_LF;
743 else if (!strcmp(value, "crlf"))
744 return EOL_CRLF;
745 return EOL_UNSET;
746}
747
92cce135 748static struct convert_driver *git_path_check_convert(struct git_attr_check *check)
aa4ed402
JH
749{
750 const char *value = check->value;
751 struct convert_driver *drv;
752
753 if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
754 return NULL;
755 for (drv = user_convert; drv; drv = drv->next)
756 if (!strcmp(value, drv->name))
757 return drv;
758 return NULL;
759}
760
92cce135 761static int git_path_check_ident(struct git_attr_check *check)
3fed15f5
JH
762{
763 const char *value = check->value;
764
765 return !!ATTR_TRUE(value);
766}
767
3bfba20d
JH
768struct conv_attrs {
769 struct convert_driver *drv;
bb211b4d
TB
770 enum crlf_action attr_action; /* What attr says */
771 enum crlf_action crlf_action; /* When no attr is set, use core.autocrlf */
3bfba20d
JH
772 int ident;
773};
774
83295964
JH
775static const char *conv_attr_name[] = {
776 "crlf", "ident", "filter", "eol", "text",
777};
778#define NUM_CONV_ATTRS ARRAY_SIZE(conv_attr_name)
779
3bfba20d 780static void convert_attrs(struct conv_attrs *ca, const char *path)
83295964
JH
781{
782 int i;
783 static struct git_attr_check ccheck[NUM_CONV_ATTRS];
784
785 if (!ccheck[0].attr) {
786 for (i = 0; i < NUM_CONV_ATTRS; i++)
787 ccheck[i].attr = git_attr(conv_attr_name[i]);
788 user_convert_tail = &user_convert;
789 git_config(read_convert_config, NULL);
790 }
3bfba20d 791
d932f4eb 792 if (!git_check_attr(path, NUM_CONV_ATTRS, ccheck)) {
bb211b4d 793 ca->crlf_action = git_path_check_crlf(ccheck + 4);
df747b81 794 if (ca->crlf_action == CRLF_UNDEFINED)
92cce135 795 ca->crlf_action = git_path_check_crlf(ccheck + 0);
bb211b4d 796 ca->attr_action = ca->crlf_action;
92cce135
TB
797 ca->ident = git_path_check_ident(ccheck + 1);
798 ca->drv = git_path_check_convert(ccheck + 2);
817a0c79
TB
799 if (ca->crlf_action != CRLF_BINARY) {
800 enum eol eol_attr = git_path_check_eol(ccheck + 3);
65237284
TB
801 if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_LF)
802 ca->crlf_action = CRLF_AUTO_INPUT;
803 else if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_CRLF)
804 ca->crlf_action = CRLF_AUTO_CRLF;
805 else if (eol_attr == EOL_LF)
817a0c79
TB
806 ca->crlf_action = CRLF_TEXT_INPUT;
807 else if (eol_attr == EOL_CRLF)
808 ca->crlf_action = CRLF_TEXT_CRLF;
809 }
810 ca->attr_action = ca->crlf_action;
3bfba20d
JH
811 } else {
812 ca->drv = NULL;
df747b81 813 ca->crlf_action = CRLF_UNDEFINED;
3bfba20d
JH
814 ca->ident = 0;
815 }
df747b81
TB
816 if (ca->crlf_action == CRLF_TEXT)
817 ca->crlf_action = text_eol_is_crlf() ? CRLF_TEXT_CRLF : CRLF_TEXT_INPUT;
818 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_FALSE)
819 ca->crlf_action = CRLF_BINARY;
820 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_TRUE)
821 ca->crlf_action = CRLF_AUTO_CRLF;
822 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_INPUT)
823 ca->crlf_action = CRLF_AUTO_INPUT;
83295964
JH
824}
825
9035d75a
SP
826int would_convert_to_git_filter_fd(const char *path)
827{
828 struct conv_attrs ca;
829
830 convert_attrs(&ca, path);
831 if (!ca.drv)
832 return 0;
833
834 /*
835 * Apply a filter to an fd only if the filter is required to succeed.
836 * We must die if the filter fails, because the original data before
837 * filtering is not available.
838 */
839 if (!ca.drv->required)
840 return 0;
841
842 return apply_filter(path, NULL, 0, -1, NULL, ca.drv->clean);
843}
844
a7630bd4
TB
845const char *get_convert_attr_ascii(const char *path)
846{
847 struct conv_attrs ca;
a7630bd4
TB
848
849 convert_attrs(&ca, path);
bb211b4d 850 switch (ca.attr_action) {
df747b81 851 case CRLF_UNDEFINED:
a7630bd4
TB
852 return "";
853 case CRLF_BINARY:
854 return "-text";
855 case CRLF_TEXT:
856 return "text";
df747b81 857 case CRLF_TEXT_INPUT:
a7630bd4 858 return "text eol=lf";
df747b81
TB
859 case CRLF_TEXT_CRLF:
860 return "text eol=crlf";
a7630bd4
TB
861 case CRLF_AUTO:
862 return "text=auto";
df747b81 863 case CRLF_AUTO_CRLF:
65237284 864 return "text=auto eol=crlf";
df747b81 865 case CRLF_AUTO_INPUT:
65237284 866 return "text=auto eol=lf";
a7630bd4
TB
867 }
868 return "";
869}
870
21e5ad50
SP
871int convert_to_git(const char *path, const char *src, size_t len,
872 struct strbuf *dst, enum safe_crlf checksafe)
35ebfd6a 873{
3bfba20d 874 int ret = 0;
cd8be6c9 875 const char *filter = NULL;
36daaaca 876 int required = 0;
3bfba20d 877 struct conv_attrs ca;
6073ee85 878
3bfba20d 879 convert_attrs(&ca, path);
36daaaca 880 if (ca.drv) {
3bfba20d 881 filter = ca.drv->clean;
36daaaca
JB
882 required = ca.drv->required;
883 }
3fed15f5 884
9035d75a 885 ret |= apply_filter(path, src, len, -1, dst, filter);
36daaaca
JB
886 if (!ret && required)
887 die("%s: clean filter '%s' failed", path, ca.drv->name);
888
92ac3197 889 if (ret && dst) {
5ecd293d
PH
890 src = dst->buf;
891 len = dst->len;
aa4ed402 892 }
3bfba20d 893 ret |= crlf_to_git(path, src, len, dst, ca.crlf_action, checksafe);
92ac3197 894 if (ret && dst) {
5ecd293d
PH
895 src = dst->buf;
896 len = dst->len;
6073ee85 897 }
3bfba20d 898 return ret | ident_to_git(path, src, len, dst, ca.ident);
35ebfd6a
JH
899}
900
9035d75a
SP
901void convert_to_git_filter_fd(const char *path, int fd, struct strbuf *dst,
902 enum safe_crlf checksafe)
903{
904 struct conv_attrs ca;
905 convert_attrs(&ca, path);
906
907 assert(ca.drv);
908 assert(ca.drv->clean);
909
910 if (!apply_filter(path, NULL, 0, fd, dst, ca.drv->clean))
911 die("%s: clean filter '%s' failed", path, ca.drv->name);
912
9035d75a
SP
913 crlf_to_git(path, dst->buf, dst->len, dst, ca.crlf_action, checksafe);
914 ident_to_git(path, dst->buf, dst->len, dst, ca.ident);
915}
916
43dd2332
EB
917static int convert_to_working_tree_internal(const char *path, const char *src,
918 size_t len, struct strbuf *dst,
919 int normalizing)
35ebfd6a 920{
36daaaca 921 int ret = 0, ret_filter = 0;
cd8be6c9 922 const char *filter = NULL;
36daaaca 923 int required = 0;
3bfba20d 924 struct conv_attrs ca;
6073ee85 925
3bfba20d 926 convert_attrs(&ca, path);
36daaaca 927 if (ca.drv) {
3bfba20d 928 filter = ca.drv->smudge;
36daaaca
JB
929 required = ca.drv->required;
930 }
3fed15f5 931
3bfba20d 932 ret |= ident_to_worktree(path, src, len, dst, ca.ident);
5ecd293d
PH
933 if (ret) {
934 src = dst->buf;
935 len = dst->len;
3fed15f5 936 }
43dd2332
EB
937 /*
938 * CRLF conversion can be skipped if normalizing, unless there
939 * is a smudge filter. The filter might expect CRLFs.
940 */
941 if (filter || !normalizing) {
3bfba20d 942 ret |= crlf_to_worktree(path, src, len, dst, ca.crlf_action);
43dd2332
EB
943 if (ret) {
944 src = dst->buf;
945 len = dst->len;
946 }
aa4ed402 947 }
36daaaca 948
9035d75a 949 ret_filter = apply_filter(path, src, len, -1, dst, filter);
36daaaca
JB
950 if (!ret_filter && required)
951 die("%s: smudge filter %s failed", path, ca.drv->name);
952
953 return ret | ret_filter;
35ebfd6a 954}
f217f0e8 955
43dd2332
EB
956int convert_to_working_tree(const char *path, const char *src, size_t len, struct strbuf *dst)
957{
958 return convert_to_working_tree_internal(path, src, len, dst, 0);
959}
960
f217f0e8
EB
961int renormalize_buffer(const char *path, const char *src, size_t len, struct strbuf *dst)
962{
43dd2332 963 int ret = convert_to_working_tree_internal(path, src, len, dst, 1);
f217f0e8
EB
964 if (ret) {
965 src = dst->buf;
966 len = dst->len;
967 }
65237284 968 return ret | convert_to_git(path, src, len, dst, SAFE_CRLF_RENORMALIZE);
f217f0e8 969}
dd8e9121 970
b6691092
JH
971/*****************************************************************
972 *
749f763d 973 * Streaming conversion support
b6691092
JH
974 *
975 *****************************************************************/
976
977typedef int (*filter_fn)(struct stream_filter *,
978 const char *input, size_t *isize_p,
979 char *output, size_t *osize_p);
980typedef void (*free_fn)(struct stream_filter *);
981
982struct stream_filter_vtbl {
983 filter_fn filter;
984 free_fn free;
985};
986
987struct stream_filter {
988 struct stream_filter_vtbl *vtbl;
989};
990
991static int null_filter_fn(struct stream_filter *filter,
992 const char *input, size_t *isize_p,
993 char *output, size_t *osize_p)
994{
4ae66704
JH
995 size_t count;
996
997 if (!input)
998 return 0; /* we do not keep any states */
999 count = *isize_p;
b6691092
JH
1000 if (*osize_p < count)
1001 count = *osize_p;
1002 if (count) {
1003 memmove(output, input, count);
1004 *isize_p -= count;
1005 *osize_p -= count;
1006 }
1007 return 0;
1008}
1009
1010static void null_free_fn(struct stream_filter *filter)
1011{
1012 ; /* nothing -- null instances are shared */
1013}
1014
1015static struct stream_filter_vtbl null_vtbl = {
1016 null_filter_fn,
1017 null_free_fn,
1018};
1019
1020static struct stream_filter null_filter_singleton = {
1021 &null_vtbl,
1022};
1023
1024int is_null_stream_filter(struct stream_filter *filter)
1025{
1026 return filter == &null_filter_singleton;
1027}
1028
b84c7839
JH
1029
1030/*
1031 * LF-to-CRLF filter
1032 */
284e3d28
CMN
1033
1034struct lf_to_crlf_filter {
1035 struct stream_filter filter;
8496f568
JH
1036 unsigned has_held:1;
1037 char held;
284e3d28
CMN
1038};
1039
e322ee38
JH
1040static int lf_to_crlf_filter_fn(struct stream_filter *filter,
1041 const char *input, size_t *isize_p,
1042 char *output, size_t *osize_p)
1043{
284e3d28
CMN
1044 size_t count, o = 0;
1045 struct lf_to_crlf_filter *lf_to_crlf = (struct lf_to_crlf_filter *)filter;
1046
8496f568
JH
1047 /*
1048 * We may be holding onto the CR to see if it is followed by a
1049 * LF, in which case we would need to go to the main loop.
1050 * Otherwise, just emit it to the output stream.
1051 */
1052 if (lf_to_crlf->has_held && (lf_to_crlf->held != '\r' || !input)) {
1053 output[o++] = lf_to_crlf->held;
1054 lf_to_crlf->has_held = 0;
284e3d28 1055 }
e322ee38 1056
87afe9a5
JH
1057 /* We are told to drain */
1058 if (!input) {
1059 *osize_p -= o;
1060 return 0;
1061 }
e322ee38 1062
e322ee38 1063 count = *isize_p;
8496f568 1064 if (count || lf_to_crlf->has_held) {
284e3d28 1065 size_t i;
8496f568
JH
1066 int was_cr = 0;
1067
1068 if (lf_to_crlf->has_held) {
1069 was_cr = 1;
1070 lf_to_crlf->has_held = 0;
1071 }
1072
284e3d28 1073 for (i = 0; o < *osize_p && i < count; i++) {
e322ee38 1074 char ch = input[i];
8496f568 1075
e322ee38 1076 if (ch == '\n') {
284e3d28 1077 output[o++] = '\r';
8496f568
JH
1078 } else if (was_cr) {
1079 /*
1080 * Previous round saw CR and it is not followed
1081 * by a LF; emit the CR before processing the
1082 * current character.
1083 */
1084 output[o++] = '\r';
e322ee38 1085 }
8496f568
JH
1086
1087 /*
1088 * We may have consumed the last output slot,
1089 * in which case we need to break out of this
1090 * loop; hold the current character before
1091 * returning.
1092 */
1093 if (*osize_p <= o) {
1094 lf_to_crlf->has_held = 1;
1095 lf_to_crlf->held = ch;
1096 continue; /* break but increment i */
1097 }
1098
1099 if (ch == '\r') {
1100 was_cr = 1;
1101 continue;
1102 }
1103
1104 was_cr = 0;
e322ee38
JH
1105 output[o++] = ch;
1106 }
1107
1108 *osize_p -= o;
1109 *isize_p -= i;
8496f568
JH
1110
1111 if (!lf_to_crlf->has_held && was_cr) {
1112 lf_to_crlf->has_held = 1;
1113 lf_to_crlf->held = '\r';
1114 }
e322ee38
JH
1115 }
1116 return 0;
1117}
1118
284e3d28
CMN
1119static void lf_to_crlf_free_fn(struct stream_filter *filter)
1120{
1121 free(filter);
1122}
1123
e322ee38
JH
1124static struct stream_filter_vtbl lf_to_crlf_vtbl = {
1125 lf_to_crlf_filter_fn,
284e3d28 1126 lf_to_crlf_free_fn,
e322ee38
JH
1127};
1128
284e3d28
CMN
1129static struct stream_filter *lf_to_crlf_filter(void)
1130{
87afe9a5 1131 struct lf_to_crlf_filter *lf_to_crlf = xcalloc(1, sizeof(*lf_to_crlf));
e322ee38 1132
284e3d28 1133 lf_to_crlf->filter.vtbl = &lf_to_crlf_vtbl;
284e3d28
CMN
1134 return (struct stream_filter *)lf_to_crlf;
1135}
b84c7839 1136
a265a7f9
JH
1137/*
1138 * Cascade filter
1139 */
1140#define FILTER_BUFFER 1024
1141struct cascade_filter {
1142 struct stream_filter filter;
1143 struct stream_filter *one;
1144 struct stream_filter *two;
1145 char buf[FILTER_BUFFER];
1146 int end, ptr;
1147};
1148
1149static int cascade_filter_fn(struct stream_filter *filter,
1150 const char *input, size_t *isize_p,
1151 char *output, size_t *osize_p)
1152{
1153 struct cascade_filter *cas = (struct cascade_filter *) filter;
1154 size_t filled = 0;
1155 size_t sz = *osize_p;
1156 size_t to_feed, remaining;
1157
1158 /*
1159 * input -- (one) --> buf -- (two) --> output
1160 */
1161 while (filled < sz) {
1162 remaining = sz - filled;
1163
1164 /* do we already have something to feed two with? */
1165 if (cas->ptr < cas->end) {
1166 to_feed = cas->end - cas->ptr;
1167 if (stream_filter(cas->two,
1168 cas->buf + cas->ptr, &to_feed,
1169 output + filled, &remaining))
1170 return -1;
1171 cas->ptr += (cas->end - cas->ptr) - to_feed;
1172 filled = sz - remaining;
1173 continue;
1174 }
1175
1176 /* feed one from upstream and have it emit into our buffer */
1177 to_feed = input ? *isize_p : 0;
1178 if (input && !to_feed)
1179 break;
1180 remaining = sizeof(cas->buf);
1181 if (stream_filter(cas->one,
1182 input, &to_feed,
1183 cas->buf, &remaining))
1184 return -1;
1185 cas->end = sizeof(cas->buf) - remaining;
1186 cas->ptr = 0;
1187 if (input) {
1188 size_t fed = *isize_p - to_feed;
1189 *isize_p -= fed;
1190 input += fed;
1191 }
1192
1193 /* do we know that we drained one completely? */
1194 if (input || cas->end)
1195 continue;
1196
1197 /* tell two to drain; we have nothing more to give it */
1198 to_feed = 0;
1199 remaining = sz - filled;
1200 if (stream_filter(cas->two,
1201 NULL, &to_feed,
1202 output + filled, &remaining))
1203 return -1;
1204 if (remaining == (sz - filled))
1205 break; /* completely drained two */
1206 filled = sz - remaining;
1207 }
1208 *osize_p -= filled;
1209 return 0;
1210}
1211
1212static void cascade_free_fn(struct stream_filter *filter)
1213{
1214 struct cascade_filter *cas = (struct cascade_filter *)filter;
1215 free_stream_filter(cas->one);
1216 free_stream_filter(cas->two);
1217 free(filter);
1218}
1219
1220static struct stream_filter_vtbl cascade_vtbl = {
1221 cascade_filter_fn,
1222 cascade_free_fn,
1223};
1224
1225static struct stream_filter *cascade_filter(struct stream_filter *one,
1226 struct stream_filter *two)
1227{
1228 struct cascade_filter *cascade;
1229
1230 if (!one || is_null_stream_filter(one))
1231 return two;
1232 if (!two || is_null_stream_filter(two))
1233 return one;
1234
1235 cascade = xmalloc(sizeof(*cascade));
1236 cascade->one = one;
1237 cascade->two = two;
1238 cascade->end = cascade->ptr = 0;
1239 cascade->filter.vtbl = &cascade_vtbl;
1240 return (struct stream_filter *)cascade;
1241}
1242
b84c7839
JH
1243/*
1244 * ident filter
1245 */
1246#define IDENT_DRAINING (-1)
1247#define IDENT_SKIPPING (-2)
1248struct ident_filter {
1249 struct stream_filter filter;
1250 struct strbuf left;
1251 int state;
1252 char ident[45]; /* ": x40 $" */
1253};
1254
1255static int is_foreign_ident(const char *str)
1256{
1257 int i;
1258
ae021d87 1259 if (!skip_prefix(str, "$Id: ", &str))
b84c7839 1260 return 0;
ae021d87 1261 for (i = 0; str[i]; i++) {
b84c7839
JH
1262 if (isspace(str[i]) && str[i+1] != '$')
1263 return 1;
1264 }
1265 return 0;
1266}
1267
1268static void ident_drain(struct ident_filter *ident, char **output_p, size_t *osize_p)
1269{
1270 size_t to_drain = ident->left.len;
1271
1272 if (*osize_p < to_drain)
1273 to_drain = *osize_p;
1274 if (to_drain) {
1275 memcpy(*output_p, ident->left.buf, to_drain);
1276 strbuf_remove(&ident->left, 0, to_drain);
1277 *output_p += to_drain;
1278 *osize_p -= to_drain;
1279 }
1280 if (!ident->left.len)
1281 ident->state = 0;
1282}
1283
1284static int ident_filter_fn(struct stream_filter *filter,
1285 const char *input, size_t *isize_p,
1286 char *output, size_t *osize_p)
1287{
1288 struct ident_filter *ident = (struct ident_filter *)filter;
1289 static const char head[] = "$Id";
1290
1291 if (!input) {
1292 /* drain upon eof */
1293 switch (ident->state) {
1294 default:
1295 strbuf_add(&ident->left, head, ident->state);
1296 case IDENT_SKIPPING:
1297 /* fallthru */
1298 case IDENT_DRAINING:
1299 ident_drain(ident, &output, osize_p);
1300 }
1301 return 0;
1302 }
1303
1304 while (*isize_p || (ident->state == IDENT_DRAINING)) {
1305 int ch;
1306
1307 if (ident->state == IDENT_DRAINING) {
1308 ident_drain(ident, &output, osize_p);
1309 if (!*osize_p)
1310 break;
1311 continue;
1312 }
1313
1314 ch = *(input++);
1315 (*isize_p)--;
1316
1317 if (ident->state == IDENT_SKIPPING) {
1318 /*
1319 * Skipping until '$' or LF, but keeping them
1320 * in case it is a foreign ident.
1321 */
1322 strbuf_addch(&ident->left, ch);
1323 if (ch != '\n' && ch != '$')
1324 continue;
1325 if (ch == '$' && !is_foreign_ident(ident->left.buf)) {
1326 strbuf_setlen(&ident->left, sizeof(head) - 1);
1327 strbuf_addstr(&ident->left, ident->ident);
1328 }
1329 ident->state = IDENT_DRAINING;
1330 continue;
1331 }
1332
1333 if (ident->state < sizeof(head) &&
1334 head[ident->state] == ch) {
1335 ident->state++;
1336 continue;
1337 }
1338
1339 if (ident->state)
1340 strbuf_add(&ident->left, head, ident->state);
1341 if (ident->state == sizeof(head) - 1) {
1342 if (ch != ':' && ch != '$') {
1343 strbuf_addch(&ident->left, ch);
1344 ident->state = 0;
1345 continue;
1346 }
1347
1348 if (ch == ':') {
1349 strbuf_addch(&ident->left, ch);
1350 ident->state = IDENT_SKIPPING;
1351 } else {
1352 strbuf_addstr(&ident->left, ident->ident);
1353 ident->state = IDENT_DRAINING;
1354 }
1355 continue;
1356 }
1357
1358 strbuf_addch(&ident->left, ch);
1359 ident->state = IDENT_DRAINING;
1360 }
1361 return 0;
1362}
1363
1364static void ident_free_fn(struct stream_filter *filter)
1365{
1366 struct ident_filter *ident = (struct ident_filter *)filter;
1367 strbuf_release(&ident->left);
1368 free(filter);
1369}
1370
1371static struct stream_filter_vtbl ident_vtbl = {
1372 ident_filter_fn,
1373 ident_free_fn,
1374};
1375
1376static struct stream_filter *ident_filter(const unsigned char *sha1)
1377{
1378 struct ident_filter *ident = xmalloc(sizeof(*ident));
1379
5096d490
JK
1380 xsnprintf(ident->ident, sizeof(ident->ident),
1381 ": %s $", sha1_to_hex(sha1));
b84c7839
JH
1382 strbuf_init(&ident->left, 0);
1383 ident->filter.vtbl = &ident_vtbl;
1384 ident->state = 0;
1385 return (struct stream_filter *)ident;
1386}
1387
dd8e9121 1388/*
b6691092
JH
1389 * Return an appropriately constructed filter for the path, or NULL if
1390 * the contents cannot be filtered without reading the whole thing
1391 * in-core.
1392 *
1393 * Note that you would be crazy to set CRLF, smuge/clean or ident to a
1394 * large binary blob you would want us not to slurp into the memory!
dd8e9121 1395 */
b6691092 1396struct stream_filter *get_stream_filter(const char *path, const unsigned char *sha1)
dd8e9121
JH
1397{
1398 struct conv_attrs ca;
b84c7839 1399 struct stream_filter *filter = NULL;
dd8e9121
JH
1400
1401 convert_attrs(&ca, path);
b84c7839 1402 if (ca.drv && (ca.drv->smudge || ca.drv->clean))
caa47adc
TB
1403 return NULL;
1404
1405 if (ca.crlf_action == CRLF_AUTO || ca.crlf_action == CRLF_AUTO_CRLF)
1406 return NULL;
b84c7839
JH
1407
1408 if (ca.ident)
1409 filter = ident_filter(sha1);
dd8e9121 1410
caa47adc 1411 if (output_eol(ca.crlf_action) == EOL_CRLF)
284e3d28 1412 filter = cascade_filter(filter, lf_to_crlf_filter());
caa47adc
TB
1413 else
1414 filter = cascade_filter(filter, &null_filter_singleton);
e322ee38 1415
b84c7839 1416 return filter;
b6691092
JH
1417}
1418
1419void free_stream_filter(struct stream_filter *filter)
1420{
1421 filter->vtbl->free(filter);
1422}
1423
1424int stream_filter(struct stream_filter *filter,
1425 const char *input, size_t *isize_p,
1426 char *output, size_t *osize_p)
1427{
1428 return filter->vtbl->filter(filter, input, isize_p, output, osize_p);
dd8e9121 1429}