?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
convert: convert convert_to_git_filter_fd to take an index
[git/debian.git] / convert.c
blobc09242cec0b1327c307b1b5a28f43f78c8e7886d
1 #include "cache.h"
2 #include "attr.h"
3 #include "run-command.h"
4 #include "quote.h"
5 #include "sigchain.h"
6 #include "pkt-line.h"
7 #include "sub-process.h"
8
9 /*
10 * convert.c - convert a file when checking it out and checking it in.
11 *
12 * This should use the pathname to decide on whether it wants to do some
13 * more interesting conversions (automatic gzip/unzip, general format
14 * conversions etc etc), but by default it just does automatic CRLF<->LF
15 * translation when the "text" attribute or "auto_crlf" option is set.
16 */
17
18 /* Stat bits: When BIN is set, the txt bits are unset */
19 #define CONVERT_STAT_BITS_TXT_LF 0x1
20 #define CONVERT_STAT_BITS_TXT_CRLF 0x2
21 #define CONVERT_STAT_BITS_BIN 0x4
22
23 enum crlf_action {
24 CRLF_UNDEFINED,
25 CRLF_BINARY,
26 CRLF_TEXT,
27 CRLF_TEXT_INPUT,
28 CRLF_TEXT_CRLF,
29 CRLF_AUTO,
30 CRLF_AUTO_INPUT,
31 CRLF_AUTO_CRLF
32 };
33
34 struct text_stat {
35 /* NUL, CR, LF and CRLF counts */
36 unsigned nul, lonecr, lonelf, crlf;
37
38 /* These are just approximations! */
39 unsigned printable, nonprintable;
40 };
41
42 static void gather_stats(const char *buf, unsigned long size, struct text_stat *stats)
43 {
44 unsigned long i;
45
46 memset(stats, 0, sizeof(*stats));
47
48 for (i = 0; i < size; i++) {
49 unsigned char c = buf[i];
50 if (c == '\r') {
51 if (i+1 < size && buf[i+1] == '\n') {
52 stats->crlf++;
53 i++;
54 } else
55 stats->lonecr++;
56 continue;
57 }
58 if (c == '\n') {
59 stats->lonelf++;
60 continue;
61 }
62 if (c == 127)
63 /* DEL */
64 stats->nonprintable++;
65 else if (c < 32) {
66 switch (c) {
67 /* BS, HT, ESC and FF */
68 case '\b': case '\t': case '\033': case '\014':
69 stats->printable++;
70 break;
71 case 0:
72 stats->nul++;
73 /* fall through */
74 default:
75 stats->nonprintable++;
76 }
77 }
78 else
79 stats->printable++;
80 }
81
82 /* If file ends with EOF then don't count this EOF as non-printable. */
83 if (size >= 1 && buf[size-1] == '\032')
84 stats->nonprintable--;
85 }
86
87 /*
88 * The same heuristics as diff.c::mmfile_is_binary()
89 * We treat files with bare CR as binary
90 */
91 static int convert_is_binary(unsigned long size, const struct text_stat *stats)
92 {
93 if (stats->lonecr)
94 return 1;
95 if (stats->nul)
96 return 1;
97 if ((stats->printable >> 7) < stats->nonprintable)
98 return 1;
99 return 0;
100 }
101
102 static unsigned int gather_convert_stats(const char *data, unsigned long size)
103 {
104 struct text_stat stats;
105 int ret = 0;
106 if (!data || !size)
107 return 0;
108 gather_stats(data, size, &stats);
109 if (convert_is_binary(size, &stats))
110 ret |= CONVERT_STAT_BITS_BIN;
111 if (stats.crlf)
112 ret |= CONVERT_STAT_BITS_TXT_CRLF;
113 if (stats.lonelf)
114 ret |= CONVERT_STAT_BITS_TXT_LF;
115
116 return ret;
117 }
118
119 static const char *gather_convert_stats_ascii(const char *data, unsigned long size)
120 {
121 unsigned int convert_stats = gather_convert_stats(data, size);
122
123 if (convert_stats & CONVERT_STAT_BITS_BIN)
124 return "-text";
125 switch (convert_stats) {
126 case CONVERT_STAT_BITS_TXT_LF:
127 return "lf";
128 case CONVERT_STAT_BITS_TXT_CRLF:
129 return "crlf";
130 case CONVERT_STAT_BITS_TXT_LF | CONVERT_STAT_BITS_TXT_CRLF:
131 return "mixed";
132 default:
133 return "none";
134 }
135 }
136
137 const char *get_cached_convert_stats_ascii(const struct index_state *istate,
138 const char *path)
139 {
140 const char *ret;
141 unsigned long sz;
142 void *data = read_blob_data_from_index(istate, path, &sz);
143 ret = gather_convert_stats_ascii(data, sz);
144 free(data);
145 return ret;
146 }
147
148 const char *get_wt_convert_stats_ascii(const char *path)
149 {
150 const char *ret = "";
151 struct strbuf sb = STRBUF_INIT;
152 if (strbuf_read_file(&sb, path, 0) >= 0)
153 ret = gather_convert_stats_ascii(sb.buf, sb.len);
154 strbuf_release(&sb);
155 return ret;
156 }
157
158 static int text_eol_is_crlf(void)
159 {
160 if (auto_crlf == AUTO_CRLF_TRUE)
161 return 1;
162 else if (auto_crlf == AUTO_CRLF_INPUT)
163 return 0;
164 if (core_eol == EOL_CRLF)
165 return 1;
166 if (core_eol == EOL_UNSET && EOL_NATIVE == EOL_CRLF)
167 return 1;
168 return 0;
169 }
170
171 static enum eol output_eol(enum crlf_action crlf_action)
172 {
173 switch (crlf_action) {
174 case CRLF_BINARY:
175 return EOL_UNSET;
176 case CRLF_TEXT_CRLF:
177 return EOL_CRLF;
178 case CRLF_TEXT_INPUT:
179 return EOL_LF;
180 case CRLF_UNDEFINED:
181 case CRLF_AUTO_CRLF:
182 return EOL_CRLF;
183 case CRLF_AUTO_INPUT:
184 return EOL_LF;
185 case CRLF_TEXT:
186 case CRLF_AUTO:
187 /* fall through */
188 return text_eol_is_crlf() ? EOL_CRLF : EOL_LF;
189 }
190 warning("Illegal crlf_action %d\n", (int)crlf_action);
191 return core_eol;
192 }
193
194 static void check_safe_crlf(const char *path, enum crlf_action crlf_action,
195 struct text_stat *old_stats, struct text_stat *new_stats,
196 enum safe_crlf checksafe)
197 {
198 if (old_stats->crlf && !new_stats->crlf ) {
199 /*
200 * CRLFs would not be restored by checkout
201 */
202 if (checksafe == SAFE_CRLF_WARN)
203 warning(_("CRLF will be replaced by LF in %s.\n"
204 "The file will have its original line"
205 " endings in your working directory."), path);
206 else /* i.e. SAFE_CRLF_FAIL */
207 die(_("CRLF would be replaced by LF in %s."), path);
208 } else if (old_stats->lonelf && !new_stats->lonelf ) {
209 /*
210 * CRLFs would be added by checkout
211 */
212 if (checksafe == SAFE_CRLF_WARN)
213 warning(_("LF will be replaced by CRLF in %s.\n"
214 "The file will have its original line"
215 " endings in your working directory."), path);
216 else /* i.e. SAFE_CRLF_FAIL */
217 die(_("LF would be replaced by CRLF in %s"), path);
218 }
219 }
220
221 static int has_cr_in_index(const struct index_state *istate, const char *path)
222 {
223 unsigned long sz;
224 void *data;
225 int has_cr;
226
227 data = read_blob_data_from_index(istate, path, &sz);
228 if (!data)
229 return 0;
230 has_cr = memchr(data, '\r', sz) != NULL;
231 free(data);
232 return has_cr;
233 }
234
235 static int will_convert_lf_to_crlf(size_t len, struct text_stat *stats,
236 enum crlf_action crlf_action)
237 {
238 if (output_eol(crlf_action) != EOL_CRLF)
239 return 0;
240 /* No "naked" LF? Nothing to convert, regardless. */
241 if (!stats->lonelf)
242 return 0;
243
244 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
245 /* If we have any CR or CRLF line endings, we do not touch it */
246 /* This is the new safer autocrlf-handling */
247 if (stats->lonecr || stats->crlf)
248 return 0;
249
250 if (convert_is_binary(len, stats))
251 return 0;
252 }
253 return 1;
254
255 }
256
257 static int crlf_to_git(const struct index_state *istate,
258 const char *path, const char *src, size_t len,
259 struct strbuf *buf,
260 enum crlf_action crlf_action, enum safe_crlf checksafe)
261 {
262 struct text_stat stats;
263 char *dst;
264 int convert_crlf_into_lf;
265
266 if (crlf_action == CRLF_BINARY ||
267 (src && !len))
268 return 0;
269
270 /*
271 * If we are doing a dry-run and have no source buffer, there is
272 * nothing to analyze; we must assume we would convert.
273 */
274 if (!buf && !src)
275 return 1;
276
277 gather_stats(src, len, &stats);
278 /* Optimization: No CRLF? Nothing to convert, regardless. */
279 convert_crlf_into_lf = !!stats.crlf;
280
281 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
282 if (convert_is_binary(len, &stats))
283 return 0;
284 /*
285 * If the file in the index has any CR in it, do not
286 * convert. This is the new safer autocrlf handling,
287 * unless we want to renormalize in a merge or
288 * cherry-pick.
289 */
290 if ((checksafe != SAFE_CRLF_RENORMALIZE) &&
291 has_cr_in_index(istate, path))
292 convert_crlf_into_lf = 0;
293 }
294 if ((checksafe == SAFE_CRLF_WARN ||
295 (checksafe == SAFE_CRLF_FAIL)) && len) {
296 struct text_stat new_stats;
297 memcpy(&new_stats, &stats, sizeof(new_stats));
298 /* simulate "git add" */
299 if (convert_crlf_into_lf) {
300 new_stats.lonelf += new_stats.crlf;
301 new_stats.crlf = 0;
302 }
303 /* simulate "git checkout" */
304 if (will_convert_lf_to_crlf(len, &new_stats, crlf_action)) {
305 new_stats.crlf += new_stats.lonelf;
306 new_stats.lonelf = 0;
307 }
308 check_safe_crlf(path, crlf_action, &stats, &new_stats, checksafe);
309 }
310 if (!convert_crlf_into_lf)
311 return 0;
312
313 /*
314 * At this point all of our source analysis is done, and we are sure we
315 * would convert. If we are in dry-run mode, we can give an answer.
316 */
317 if (!buf)
318 return 1;
319
320 /* only grow if not in place */
321 if (strbuf_avail(buf) + buf->len < len)
322 strbuf_grow(buf, len - buf->len);
323 dst = buf->buf;
324 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
325 /*
326 * If we guessed, we already know we rejected a file with
327 * lone CR, and we can strip a CR without looking at what
328 * follow it.
329 */
330 do {
331 unsigned char c = *src++;
332 if (c != '\r')
333 *dst++ = c;
334 } while (--len);
335 } else {
336 do {
337 unsigned char c = *src++;
338 if (! (c == '\r' && (1 < len && *src == '\n')))
339 *dst++ = c;
340 } while (--len);
341 }
342 strbuf_setlen(buf, dst - buf->buf);
343 return 1;
344 }
345
346 static int crlf_to_worktree(const char *path, const char *src, size_t len,
347 struct strbuf *buf, enum crlf_action crlf_action)
348 {
349 char *to_free = NULL;
350 struct text_stat stats;
351
352 if (!len || output_eol(crlf_action) != EOL_CRLF)
353 return 0;
354
355 gather_stats(src, len, &stats);
356 if (!will_convert_lf_to_crlf(len, &stats, crlf_action))
357 return 0;
358
359 /* are we "faking" in place editing ? */
360 if (src == buf->buf)
361 to_free = strbuf_detach(buf, NULL);
362
363 strbuf_grow(buf, len + stats.lonelf);
364 for (;;) {
365 const char *nl = memchr(src, '\n', len);
366 if (!nl)
367 break;
368 if (nl > src && nl[-1] == '\r') {
369 strbuf_add(buf, src, nl + 1 - src);
370 } else {
371 strbuf_add(buf, src, nl - src);
372 strbuf_addstr(buf, "\r\n");
373 }
374 len -= nl + 1 - src;
375 src = nl + 1;
376 }
377 strbuf_add(buf, src, len);
378
379 free(to_free);
380 return 1;
381 }
382
383 struct filter_params {
384 const char *src;
385 unsigned long size;
386 int fd;
387 const char *cmd;
388 const char *path;
389 };
390
391 static int filter_buffer_or_fd(int in, int out, void *data)
392 {
393 /*
394 * Spawn cmd and feed the buffer contents through its stdin.
395 */
396 struct child_process child_process = CHILD_PROCESS_INIT;
397 struct filter_params *params = (struct filter_params *)data;
398 int write_err, status;
399 const char *argv[] = { NULL, NULL };
400
401 /* apply % substitution to cmd */
402 struct strbuf cmd = STRBUF_INIT;
403 struct strbuf path = STRBUF_INIT;
404 struct strbuf_expand_dict_entry dict[] = {
405 { "f", NULL, },
406 { NULL, NULL, },
407 };
408
409 /* quote the path to preserve spaces, etc. */
410 sq_quote_buf(&path, params->path);
411 dict[0].value = path.buf;
412
413 /* expand all %f with the quoted path */
414 strbuf_expand(&cmd, params->cmd, strbuf_expand_dict_cb, &dict);
415 strbuf_release(&path);
416
417 argv[0] = cmd.buf;
418
419 child_process.argv = argv;
420 child_process.use_shell = 1;
421 child_process.in = -1;
422 child_process.out = out;
423
424 if (start_command(&child_process))
425 return error("cannot fork to run external filter '%s'", params->cmd);
426
427 sigchain_push(SIGPIPE, SIG_IGN);
428
429 if (params->src) {
430 write_err = (write_in_full(child_process.in,
431 params->src, params->size) < 0);
432 if (errno == EPIPE)
433 write_err = 0;
434 } else {
435 write_err = copy_fd(params->fd, child_process.in);
436 if (write_err == COPY_WRITE_ERROR && errno == EPIPE)
437 write_err = 0;
438 }
439
440 if (close(child_process.in))
441 write_err = 1;
442 if (write_err)
443 error("cannot feed the input to external filter '%s'", params->cmd);
444
445 sigchain_pop(SIGPIPE);
446
447 status = finish_command(&child_process);
448 if (status)
449 error("external filter '%s' failed %d", params->cmd, status);
450
451 strbuf_release(&cmd);
452 return (write_err || status);
453 }
454
455 static int apply_single_file_filter(const char *path, const char *src, size_t len, int fd,
456 struct strbuf *dst, const char *cmd)
457 {
458 /*
459 * Create a pipeline to have the command filter the buffer's
460 * contents.
461 *
462 * (child --> cmd) --> us
463 */
464 int err = 0;
465 struct strbuf nbuf = STRBUF_INIT;
466 struct async async;
467 struct filter_params params;
468
469 memset(&async, 0, sizeof(async));
470 async.proc = filter_buffer_or_fd;
471 async.data = &params;
472 async.out = -1;
473 params.src = src;
474 params.size = len;
475 params.fd = fd;
476 params.cmd = cmd;
477 params.path = path;
478
479 fflush(NULL);
480 if (start_async(&async))
481 return 0; /* error was already reported */
482
483 if (strbuf_read(&nbuf, async.out, len) < 0) {
484 err = error("read from external filter '%s' failed", cmd);
485 }
486 if (close(async.out)) {
487 err = error("read from external filter '%s' failed", cmd);
488 }
489 if (finish_async(&async)) {
490 err = error("external filter '%s' failed", cmd);
491 }
492
493 if (!err) {
494 strbuf_swap(dst, &nbuf);
495 }
496 strbuf_release(&nbuf);
497 return !err;
498 }
499
500 #define CAP_CLEAN (1u<<0)
501 #define CAP_SMUDGE (1u<<1)
502
503 struct cmd2process {
504 struct subprocess_entry subprocess; /* must be the first member! */
505 unsigned int supported_capabilities;
506 };
507
508 static int subprocess_map_initialized;
509 static struct hashmap subprocess_map;
510
511 static int start_multi_file_filter_fn(struct subprocess_entry *subprocess)
512 {
513 int err;
514 struct cmd2process *entry = (struct cmd2process *)subprocess;
515 struct string_list cap_list = STRING_LIST_INIT_NODUP;
516 char *cap_buf;
517 const char *cap_name;
518 struct child_process *process = &subprocess->process;
519 const char *cmd = subprocess->cmd;
520
521 sigchain_push(SIGPIPE, SIG_IGN);
522
523 err = packet_writel(process->in, "git-filter-client", "version=2", NULL);
524 if (err)
525 goto done;
526
527 err = strcmp(packet_read_line(process->out, NULL), "git-filter-server");
528 if (err) {
529 error("external filter '%s' does not support filter protocol version 2", cmd);
530 goto done;
531 }
532 err = strcmp(packet_read_line(process->out, NULL), "version=2");
533 if (err)
534 goto done;
535 err = packet_read_line(process->out, NULL) != NULL;
536 if (err)
537 goto done;
538
539 err = packet_writel(process->in, "capability=clean", "capability=smudge", NULL);
540
541 for (;;) {
542 cap_buf = packet_read_line(process->out, NULL);
543 if (!cap_buf)
544 break;
545 string_list_split_in_place(&cap_list, cap_buf, '=', 1);
546
547 if (cap_list.nr != 2 || strcmp(cap_list.items[0].string, "capability"))
548 continue;
549
550 cap_name = cap_list.items[1].string;
551 if (!strcmp(cap_name, "clean")) {
552 entry->supported_capabilities |= CAP_CLEAN;
553 } else if (!strcmp(cap_name, "smudge")) {
554 entry->supported_capabilities |= CAP_SMUDGE;
555 } else {
556 warning(
557 "external filter '%s' requested unsupported filter capability '%s'",
558 cmd, cap_name
559 );
560 }
561
562 string_list_clear(&cap_list, 0);
563 }
564
565 done:
566 sigchain_pop(SIGPIPE);
567
568 return err;
569 }
570
571 static int apply_multi_file_filter(const char *path, const char *src, size_t len,
572 int fd, struct strbuf *dst, const char *cmd,
573 const unsigned int wanted_capability)
574 {
575 int err;
576 struct cmd2process *entry;
577 struct child_process *process;
578 struct strbuf nbuf = STRBUF_INIT;
579 struct strbuf filter_status = STRBUF_INIT;
580 const char *filter_type;
581
582 if (!subprocess_map_initialized) {
583 subprocess_map_initialized = 1;
584 hashmap_init(&subprocess_map, (hashmap_cmp_fn) cmd2process_cmp, 0);
585 entry = NULL;
586 } else {
587 entry = (struct cmd2process *)subprocess_find_entry(&subprocess_map, cmd);
588 }
589
590 fflush(NULL);
591
592 if (!entry) {
593 entry = xmalloc(sizeof(*entry));
594 entry->supported_capabilities = 0;
595
596 if (subprocess_start(&subprocess_map, &entry->subprocess, cmd, start_multi_file_filter_fn)) {
597 free(entry);
598 return 0;
599 }
600 }
601 process = &entry->subprocess.process;
602
603 if (!(wanted_capability & entry->supported_capabilities))
604 return 0;
605
606 if (CAP_CLEAN & wanted_capability)
607 filter_type = "clean";
608 else if (CAP_SMUDGE & wanted_capability)
609 filter_type = "smudge";
610 else
611 die("unexpected filter type");
612
613 sigchain_push(SIGPIPE, SIG_IGN);
614
615 assert(strlen(filter_type) < LARGE_PACKET_DATA_MAX - strlen("command=\n"));
616 err = packet_write_fmt_gently(process->in, "command=%s\n", filter_type);
617 if (err)
618 goto done;
619
620 err = strlen(path) > LARGE_PACKET_DATA_MAX - strlen("pathname=\n");
621 if (err) {
622 error("path name too long for external filter");
623 goto done;
624 }
625
626 err = packet_write_fmt_gently(process->in, "pathname=%s\n", path);
627 if (err)
628 goto done;
629
630 err = packet_flush_gently(process->in);
631 if (err)
632 goto done;
633
634 if (fd >= 0)
635 err = write_packetized_from_fd(fd, process->in);
636 else
637 err = write_packetized_from_buf(src, len, process->in);
638 if (err)
639 goto done;
640
641 err = subprocess_read_status(process->out, &filter_status);
642 if (err)
643 goto done;
644
645 err = strcmp(filter_status.buf, "success");
646 if (err)
647 goto done;
648
649 err = read_packetized_to_strbuf(process->out, &nbuf) < 0;
650 if (err)
651 goto done;
652
653 err = subprocess_read_status(process->out, &filter_status);
654 if (err)
655 goto done;
656
657 err = strcmp(filter_status.buf, "success");
658
659 done:
660 sigchain_pop(SIGPIPE);
661
662 if (err) {
663 if (!strcmp(filter_status.buf, "error")) {
664 /* The filter signaled a problem with the file. */
665 } else if (!strcmp(filter_status.buf, "abort")) {
666 /*
667 * The filter signaled a permanent problem. Don't try to filter
668 * files with the same command for the lifetime of the current
669 * Git process.
670 */
671 entry->supported_capabilities &= ~wanted_capability;
672 } else {
673 /*
674 * Something went wrong with the protocol filter.
675 * Force shutdown and restart if another blob requires filtering.
676 */
677 error("external filter '%s' failed", cmd);
678 subprocess_stop(&subprocess_map, &entry->subprocess);
679 free(entry);
680 }
681 } else {
682 strbuf_swap(dst, &nbuf);
683 }
684 strbuf_release(&nbuf);
685 return !err;
686 }
687
688 static struct convert_driver {
689 const char *name;
690 struct convert_driver *next;
691 const char *smudge;
692 const char *clean;
693 const char *process;
694 int required;
695 } *user_convert, **user_convert_tail;
696
697 static int apply_filter(const char *path, const char *src, size_t len,
698 int fd, struct strbuf *dst, struct convert_driver *drv,
699 const unsigned int wanted_capability)
700 {
701 const char *cmd = NULL;
702
703 if (!drv)
704 return 0;
705
706 if (!dst)
707 return 1;
708
709 if ((CAP_CLEAN & wanted_capability) && !drv->process && drv->clean)
710 cmd = drv->clean;
711 else if ((CAP_SMUDGE & wanted_capability) && !drv->process && drv->smudge)
712 cmd = drv->smudge;
713
714 if (cmd && *cmd)
715 return apply_single_file_filter(path, src, len, fd, dst, cmd);
716 else if (drv->process && *drv->process)
717 return apply_multi_file_filter(path, src, len, fd, dst, drv->process, wanted_capability);
718
719 return 0;
720 }
721
722 static int read_convert_config(const char *var, const char *value, void *cb)
723 {
724 const char *key, *name;
725 int namelen;
726 struct convert_driver *drv;
727
728 /*
729 * External conversion drivers are configured using
730 * "filter.<name>.variable".
731 */
732 if (parse_config_key(var, "filter", &name, &namelen, &key) < 0 || !name)
733 return 0;
734 for (drv = user_convert; drv; drv = drv->next)
735 if (!strncmp(drv->name, name, namelen) && !drv->name[namelen])
736 break;
737 if (!drv) {
738 drv = xcalloc(1, sizeof(struct convert_driver));
739 drv->name = xmemdupz(name, namelen);
740 *user_convert_tail = drv;
741 user_convert_tail = &(drv->next);
742 }
743
744 /*
745 * filter.<name>.smudge and filter.<name>.clean specifies
746 * the command line:
747 *
748 * command-line
749 *
750 * The command-line will not be interpolated in any way.
751 */
752
753 if (!strcmp("smudge", key))
754 return git_config_string(&drv->smudge, var, value);
755
756 if (!strcmp("clean", key))
757 return git_config_string(&drv->clean, var, value);
758
759 if (!strcmp("process", key))
760 return git_config_string(&drv->process, var, value);
761
762 if (!strcmp("required", key)) {
763 drv->required = git_config_bool(var, value);
764 return 0;
765 }
766
767 return 0;
768 }
769
770 static int count_ident(const char *cp, unsigned long size)
771 {
772 /*
773 * "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$"
774 */
775 int cnt = 0;
776 char ch;
777
778 while (size) {
779 ch = *cp++;
780 size--;
781 if (ch != '$')
782 continue;
783 if (size < 3)
784 break;
785 if (memcmp("Id", cp, 2))
786 continue;
787 ch = cp[2];
788 cp += 3;
789 size -= 3;
790 if (ch == '$')
791 cnt++; /* $Id$ */
792 if (ch != ':')
793 continue;
794
795 /*
796 * "$Id: ... "; scan up to the closing dollar sign and discard.
797 */
798 while (size) {
799 ch = *cp++;
800 size--;
801 if (ch == '$') {
802 cnt++;
803 break;
804 }
805 if (ch == '\n')
806 break;
807 }
808 }
809 return cnt;
810 }
811
812 static int ident_to_git(const char *path, const char *src, size_t len,
813 struct strbuf *buf, int ident)
814 {
815 char *dst, *dollar;
816
817 if (!ident || (src && !count_ident(src, len)))
818 return 0;
819
820 if (!buf)
821 return 1;
822
823 /* only grow if not in place */
824 if (strbuf_avail(buf) + buf->len < len)
825 strbuf_grow(buf, len - buf->len);
826 dst = buf->buf;
827 for (;;) {
828 dollar = memchr(src, '$', len);
829 if (!dollar)
830 break;
831 memmove(dst, src, dollar + 1 - src);
832 dst += dollar + 1 - src;
833 len -= dollar + 1 - src;
834 src = dollar + 1;
835
836 if (len > 3 && !memcmp(src, "Id:", 3)) {
837 dollar = memchr(src + 3, '$', len - 3);
838 if (!dollar)
839 break;
840 if (memchr(src + 3, '\n', dollar - src - 3)) {
841 /* Line break before the next dollar. */
842 continue;
843 }
844
845 memcpy(dst, "Id$", 3);
846 dst += 3;
847 len -= dollar + 1 - src;
848 src = dollar + 1;
849 }
850 }
851 memmove(dst, src, len);
852 strbuf_setlen(buf, dst + len - buf->buf);
853 return 1;
854 }
855
856 static int ident_to_worktree(const char *path, const char *src, size_t len,
857 struct strbuf *buf, int ident)
858 {
859 unsigned char sha1[20];
860 char *to_free = NULL, *dollar, *spc;
861 int cnt;
862
863 if (!ident)
864 return 0;
865
866 cnt = count_ident(src, len);
867 if (!cnt)
868 return 0;
869
870 /* are we "faking" in place editing ? */
871 if (src == buf->buf)
872 to_free = strbuf_detach(buf, NULL);
873 hash_sha1_file(src, len, "blob", sha1);
874
875 strbuf_grow(buf, len + cnt * 43);
876 for (;;) {
877 /* step 1: run to the next '$' */
878 dollar = memchr(src, '$', len);
879 if (!dollar)
880 break;
881 strbuf_add(buf, src, dollar + 1 - src);
882 len -= dollar + 1 - src;
883 src = dollar + 1;
884
885 /* step 2: does it looks like a bit like Id:xxx$ or Id$ ? */
886 if (len < 3 || memcmp("Id", src, 2))
887 continue;
888
889 /* step 3: skip over Id$ or Id:xxxxx$ */
890 if (src[2] == '$') {
891 src += 3;
892 len -= 3;
893 } else if (src[2] == ':') {
894 /*
895 * It's possible that an expanded Id has crept its way into the
896 * repository, we cope with that by stripping the expansion out.
897 * This is probably not a good idea, since it will cause changes
898 * on checkout, which won't go away by stash, but let's keep it
899 * for git-style ids.
900 */
901 dollar = memchr(src + 3, '$', len - 3);
902 if (!dollar) {
903 /* incomplete keyword, no more '$', so just quit the loop */
904 break;
905 }
906
907 if (memchr(src + 3, '\n', dollar - src - 3)) {
908 /* Line break before the next dollar. */
909 continue;
910 }
911
912 spc = memchr(src + 4, ' ', dollar - src - 4);
913 if (spc && spc < dollar-1) {
914 /* There are spaces in unexpected places.
915 * This is probably an id from some other
916 * versioning system. Keep it for now.
917 */
918 continue;
919 }
920
921 len -= dollar + 1 - src;
922 src = dollar + 1;
923 } else {
924 /* it wasn't a "Id$" or "Id:xxxx$" */
925 continue;
926 }
927
928 /* step 4: substitute */
929 strbuf_addstr(buf, "Id: ");
930 strbuf_add(buf, sha1_to_hex(sha1), 40);
931 strbuf_addstr(buf, " $");
932 }
933 strbuf_add(buf, src, len);
934
935 free(to_free);
936 return 1;
937 }
938
939 static enum crlf_action git_path_check_crlf(struct attr_check_item *check)
940 {
941 const char *value = check->value;
942
943 if (ATTR_TRUE(value))
944 return CRLF_TEXT;
945 else if (ATTR_FALSE(value))
946 return CRLF_BINARY;
947 else if (ATTR_UNSET(value))
948 ;
949 else if (!strcmp(value, "input"))
950 return CRLF_TEXT_INPUT;
951 else if (!strcmp(value, "auto"))
952 return CRLF_AUTO;
953 return CRLF_UNDEFINED;
954 }
955
956 static enum eol git_path_check_eol(struct attr_check_item *check)
957 {
958 const char *value = check->value;
959
960 if (ATTR_UNSET(value))
961 ;
962 else if (!strcmp(value, "lf"))
963 return EOL_LF;
964 else if (!strcmp(value, "crlf"))
965 return EOL_CRLF;
966 return EOL_UNSET;
967 }
968
969 static struct convert_driver *git_path_check_convert(struct attr_check_item *check)
970 {
971 const char *value = check->value;
972 struct convert_driver *drv;
973
974 if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
975 return NULL;
976 for (drv = user_convert; drv; drv = drv->next)
977 if (!strcmp(value, drv->name))
978 return drv;
979 return NULL;
980 }
981
982 static int git_path_check_ident(struct attr_check_item *check)
983 {
984 const char *value = check->value;
985
986 return !!ATTR_TRUE(value);
987 }
988
989 struct conv_attrs {
990 struct convert_driver *drv;
991 enum crlf_action attr_action; /* What attr says */
992 enum crlf_action crlf_action; /* When no attr is set, use core.autocrlf */
993 int ident;
994 };
995
996 static void convert_attrs(struct conv_attrs *ca, const char *path)
997 {
998 static struct attr_check *check;
999
1000 if (!check) {
1001 check = attr_check_initl("crlf", "ident", "filter",
1002 "eol", "text", NULL);
1003 user_convert_tail = &user_convert;
1004 git_config(read_convert_config, NULL);
1005 }
1006
1007 if (!git_check_attr(path, check)) {
1008 struct attr_check_item *ccheck = check->items;
1009 ca->crlf_action = git_path_check_crlf(ccheck + 4);
1010 if (ca->crlf_action == CRLF_UNDEFINED)
1011 ca->crlf_action = git_path_check_crlf(ccheck + 0);
1012 ca->attr_action = ca->crlf_action;
1013 ca->ident = git_path_check_ident(ccheck + 1);
1014 ca->drv = git_path_check_convert(ccheck + 2);
1015 if (ca->crlf_action != CRLF_BINARY) {
1016 enum eol eol_attr = git_path_check_eol(ccheck + 3);
1017 if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_LF)
1018 ca->crlf_action = CRLF_AUTO_INPUT;
1019 else if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_CRLF)
1020 ca->crlf_action = CRLF_AUTO_CRLF;
1021 else if (eol_attr == EOL_LF)
1022 ca->crlf_action = CRLF_TEXT_INPUT;
1023 else if (eol_attr == EOL_CRLF)
1024 ca->crlf_action = CRLF_TEXT_CRLF;
1025 }
1026 ca->attr_action = ca->crlf_action;
1027 } else {
1028 ca->drv = NULL;
1029 ca->crlf_action = CRLF_UNDEFINED;
1030 ca->ident = 0;
1031 }
1032 if (ca->crlf_action == CRLF_TEXT)
1033 ca->crlf_action = text_eol_is_crlf() ? CRLF_TEXT_CRLF : CRLF_TEXT_INPUT;
1034 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_FALSE)
1035 ca->crlf_action = CRLF_BINARY;
1036 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_TRUE)
1037 ca->crlf_action = CRLF_AUTO_CRLF;
1038 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_INPUT)
1039 ca->crlf_action = CRLF_AUTO_INPUT;
1040 }
1041
1042 int would_convert_to_git_filter_fd(const char *path)
1043 {
1044 struct conv_attrs ca;
1045
1046 convert_attrs(&ca, path);
1047 if (!ca.drv)
1048 return 0;
1049
1050 /*
1051 * Apply a filter to an fd only if the filter is required to succeed.
1052 * We must die if the filter fails, because the original data before
1053 * filtering is not available.
1054 */
1055 if (!ca.drv->required)
1056 return 0;
1057
1058 return apply_filter(path, NULL, 0, -1, NULL, ca.drv, CAP_CLEAN);
1059 }
1060
1061 const char *get_convert_attr_ascii(const char *path)
1062 {
1063 struct conv_attrs ca;
1064
1065 convert_attrs(&ca, path);
1066 switch (ca.attr_action) {
1067 case CRLF_UNDEFINED:
1068 return "";
1069 case CRLF_BINARY:
1070 return "-text";
1071 case CRLF_TEXT:
1072 return "text";
1073 case CRLF_TEXT_INPUT:
1074 return "text eol=lf";
1075 case CRLF_TEXT_CRLF:
1076 return "text eol=crlf";
1077 case CRLF_AUTO:
1078 return "text=auto";
1079 case CRLF_AUTO_CRLF:
1080 return "text=auto eol=crlf";
1081 case CRLF_AUTO_INPUT:
1082 return "text=auto eol=lf";
1083 }
1084 return "";
1085 }
1086
1087 int convert_to_git(const char *path, const char *src, size_t len,
1088 struct strbuf *dst, enum safe_crlf checksafe)
1089 {
1090 int ret = 0;
1091 struct conv_attrs ca;
1092
1093 convert_attrs(&ca, path);
1094
1095 ret |= apply_filter(path, src, len, -1, dst, ca.drv, CAP_CLEAN);
1096 if (!ret && ca.drv && ca.drv->required)
1097 die("%s: clean filter '%s' failed", path, ca.drv->name);
1098
1099 if (ret && dst) {
1100 src = dst->buf;
1101 len = dst->len;
1102 }
1103 ret |= crlf_to_git(&the_index, path, src, len, dst, ca.crlf_action, checksafe);
1104 if (ret && dst) {
1105 src = dst->buf;
1106 len = dst->len;
1107 }
1108 return ret | ident_to_git(path, src, len, dst, ca.ident);
1109 }
1110
1111 void convert_to_git_filter_fd(const struct index_state *istate,
1112 const char *path, int fd, struct strbuf *dst,
1113 enum safe_crlf checksafe)
1114 {
1115 struct conv_attrs ca;
1116 convert_attrs(&ca, path);
1117
1118 assert(ca.drv);
1119 assert(ca.drv->clean || ca.drv->process);
1120
1121 if (!apply_filter(path, NULL, 0, fd, dst, ca.drv, CAP_CLEAN))
1122 die("%s: clean filter '%s' failed", path, ca.drv->name);
1123
1124 crlf_to_git(istate, path, dst->buf, dst->len, dst, ca.crlf_action, checksafe);
1125 ident_to_git(path, dst->buf, dst->len, dst, ca.ident);
1126 }
1127
1128 static int convert_to_working_tree_internal(const char *path, const char *src,
1129 size_t len, struct strbuf *dst,
1130 int normalizing)
1131 {
1132 int ret = 0, ret_filter = 0;
1133 struct conv_attrs ca;
1134
1135 convert_attrs(&ca, path);
1136
1137 ret |= ident_to_worktree(path, src, len, dst, ca.ident);
1138 if (ret) {
1139 src = dst->buf;
1140 len = dst->len;
1141 }
1142 /*
1143 * CRLF conversion can be skipped if normalizing, unless there
1144 * is a smudge or process filter (even if the process filter doesn't
1145 * support smudge). The filters might expect CRLFs.
1146 */
1147 if ((ca.drv && (ca.drv->smudge || ca.drv->process)) || !normalizing) {
1148 ret |= crlf_to_worktree(path, src, len, dst, ca.crlf_action);
1149 if (ret) {
1150 src = dst->buf;
1151 len = dst->len;
1152 }
1153 }
1154
1155 ret_filter = apply_filter(path, src, len, -1, dst, ca.drv, CAP_SMUDGE);
1156 if (!ret_filter && ca.drv && ca.drv->required)
1157 die("%s: smudge filter %s failed", path, ca.drv->name);
1158
1159 return ret | ret_filter;
1160 }
1161
1162 int convert_to_working_tree(const char *path, const char *src, size_t len, struct strbuf *dst)
1163 {
1164 return convert_to_working_tree_internal(path, src, len, dst, 0);
1165 }
1166
1167 int renormalize_buffer(const char *path, const char *src, size_t len, struct strbuf *dst)
1168 {
1169 int ret = convert_to_working_tree_internal(path, src, len, dst, 1);
1170 if (ret) {
1171 src = dst->buf;
1172 len = dst->len;
1173 }
1174 return ret | convert_to_git(path, src, len, dst, SAFE_CRLF_RENORMALIZE);
1175 }
1176
1177 /*****************************************************************
1178 *
1179 * Streaming conversion support
1180 *
1181 *****************************************************************/
1182
1183 typedef int (*filter_fn)(struct stream_filter *,
1184 const char *input, size_t *isize_p,
1185 char *output, size_t *osize_p);
1186 typedef void (*free_fn)(struct stream_filter *);
1187
1188 struct stream_filter_vtbl {
1189 filter_fn filter;
1190 free_fn free;
1191 };
1192
1193 struct stream_filter {
1194 struct stream_filter_vtbl *vtbl;
1195 };
1196
1197 static int null_filter_fn(struct stream_filter *filter,
1198 const char *input, size_t *isize_p,
1199 char *output, size_t *osize_p)
1200 {
1201 size_t count;
1202
1203 if (!input)
1204 return 0; /* we do not keep any states */
1205 count = *isize_p;
1206 if (*osize_p < count)
1207 count = *osize_p;
1208 if (count) {
1209 memmove(output, input, count);
1210 *isize_p -= count;
1211 *osize_p -= count;
1212 }
1213 return 0;
1214 }
1215
1216 static void null_free_fn(struct stream_filter *filter)
1217 {
1218 ; /* nothing -- null instances are shared */
1219 }
1220
1221 static struct stream_filter_vtbl null_vtbl = {
1222 null_filter_fn,
1223 null_free_fn,
1224 };
1225
1226 static struct stream_filter null_filter_singleton = {
1227 &null_vtbl,
1228 };
1229
1230 int is_null_stream_filter(struct stream_filter *filter)
1231 {
1232 return filter == &null_filter_singleton;
1233 }
1234
1235
1236 /*
1237 * LF-to-CRLF filter
1238 */
1239
1240 struct lf_to_crlf_filter {
1241 struct stream_filter filter;
1242 unsigned has_held:1;
1243 char held;
1244 };
1245
1246 static int lf_to_crlf_filter_fn(struct stream_filter *filter,
1247 const char *input, size_t *isize_p,
1248 char *output, size_t *osize_p)
1249 {
1250 size_t count, o = 0;
1251 struct lf_to_crlf_filter *lf_to_crlf = (struct lf_to_crlf_filter *)filter;
1252
1253 /*
1254 * We may be holding onto the CR to see if it is followed by a
1255 * LF, in which case we would need to go to the main loop.
1256 * Otherwise, just emit it to the output stream.
1257 */
1258 if (lf_to_crlf->has_held && (lf_to_crlf->held != '\r' || !input)) {
1259 output[o++] = lf_to_crlf->held;
1260 lf_to_crlf->has_held = 0;
1261 }
1262
1263 /* We are told to drain */
1264 if (!input) {
1265 *osize_p -= o;
1266 return 0;
1267 }
1268
1269 count = *isize_p;
1270 if (count || lf_to_crlf->has_held) {
1271 size_t i;
1272 int was_cr = 0;
1273
1274 if (lf_to_crlf->has_held) {
1275 was_cr = 1;
1276 lf_to_crlf->has_held = 0;
1277 }
1278
1279 for (i = 0; o < *osize_p && i < count; i++) {
1280 char ch = input[i];
1281
1282 if (ch == '\n') {
1283 output[o++] = '\r';
1284 } else if (was_cr) {
1285 /*
1286 * Previous round saw CR and it is not followed
1287 * by a LF; emit the CR before processing the
1288 * current character.
1289 */
1290 output[o++] = '\r';
1291 }
1292
1293 /*
1294 * We may have consumed the last output slot,
1295 * in which case we need to break out of this
1296 * loop; hold the current character before
1297 * returning.
1298 */
1299 if (*osize_p <= o) {
1300 lf_to_crlf->has_held = 1;
1301 lf_to_crlf->held = ch;
1302 continue; /* break but increment i */
1303 }
1304
1305 if (ch == '\r') {
1306 was_cr = 1;
1307 continue;
1308 }
1309
1310 was_cr = 0;
1311 output[o++] = ch;
1312 }
1313
1314 *osize_p -= o;
1315 *isize_p -= i;
1316
1317 if (!lf_to_crlf->has_held && was_cr) {
1318 lf_to_crlf->has_held = 1;
1319 lf_to_crlf->held = '\r';
1320 }
1321 }
1322 return 0;
1323 }
1324
1325 static void lf_to_crlf_free_fn(struct stream_filter *filter)
1326 {
1327 free(filter);
1328 }
1329
1330 static struct stream_filter_vtbl lf_to_crlf_vtbl = {
1331 lf_to_crlf_filter_fn,
1332 lf_to_crlf_free_fn,
1333 };
1334
1335 static struct stream_filter *lf_to_crlf_filter(void)
1336 {
1337 struct lf_to_crlf_filter *lf_to_crlf = xcalloc(1, sizeof(*lf_to_crlf));
1338
1339 lf_to_crlf->filter.vtbl = &lf_to_crlf_vtbl;
1340 return (struct stream_filter *)lf_to_crlf;
1341 }
1342
1343 /*
1344 * Cascade filter
1345 */
1346 #define FILTER_BUFFER 1024
1347 struct cascade_filter {
1348 struct stream_filter filter;
1349 struct stream_filter *one;
1350 struct stream_filter *two;
1351 char buf[FILTER_BUFFER];
1352 int end, ptr;
1353 };
1354
1355 static int cascade_filter_fn(struct stream_filter *filter,
1356 const char *input, size_t *isize_p,
1357 char *output, size_t *osize_p)
1358 {
1359 struct cascade_filter *cas = (struct cascade_filter *) filter;
1360 size_t filled = 0;
1361 size_t sz = *osize_p;
1362 size_t to_feed, remaining;
1363
1364 /*
1365 * input -- (one) --> buf -- (two) --> output
1366 */
1367 while (filled < sz) {
1368 remaining = sz - filled;
1369
1370 /* do we already have something to feed two with? */
1371 if (cas->ptr < cas->end) {
1372 to_feed = cas->end - cas->ptr;
1373 if (stream_filter(cas->two,
1374 cas->buf + cas->ptr, &to_feed,
1375 output + filled, &remaining))
1376 return -1;
1377 cas->ptr += (cas->end - cas->ptr) - to_feed;
1378 filled = sz - remaining;
1379 continue;
1380 }
1381
1382 /* feed one from upstream and have it emit into our buffer */
1383 to_feed = input ? *isize_p : 0;
1384 if (input && !to_feed)
1385 break;
1386 remaining = sizeof(cas->buf);
1387 if (stream_filter(cas->one,
1388 input, &to_feed,
1389 cas->buf, &remaining))
1390 return -1;
1391 cas->end = sizeof(cas->buf) - remaining;
1392 cas->ptr = 0;
1393 if (input) {
1394 size_t fed = *isize_p - to_feed;
1395 *isize_p -= fed;
1396 input += fed;
1397 }
1398
1399 /* do we know that we drained one completely? */
1400 if (input || cas->end)
1401 continue;
1402
1403 /* tell two to drain; we have nothing more to give it */
1404 to_feed = 0;
1405 remaining = sz - filled;
1406 if (stream_filter(cas->two,
1407 NULL, &to_feed,
1408 output + filled, &remaining))
1409 return -1;
1410 if (remaining == (sz - filled))
1411 break; /* completely drained two */
1412 filled = sz - remaining;
1413 }
1414 *osize_p -= filled;
1415 return 0;
1416 }
1417
1418 static void cascade_free_fn(struct stream_filter *filter)
1419 {
1420 struct cascade_filter *cas = (struct cascade_filter *)filter;
1421 free_stream_filter(cas->one);
1422 free_stream_filter(cas->two);
1423 free(filter);
1424 }
1425
1426 static struct stream_filter_vtbl cascade_vtbl = {
1427 cascade_filter_fn,
1428 cascade_free_fn,
1429 };
1430
1431 static struct stream_filter *cascade_filter(struct stream_filter *one,
1432 struct stream_filter *two)
1433 {
1434 struct cascade_filter *cascade;
1435
1436 if (!one || is_null_stream_filter(one))
1437 return two;
1438 if (!two || is_null_stream_filter(two))
1439 return one;
1440
1441 cascade = xmalloc(sizeof(*cascade));
1442 cascade->one = one;
1443 cascade->two = two;
1444 cascade->end = cascade->ptr = 0;
1445 cascade->filter.vtbl = &cascade_vtbl;
1446 return (struct stream_filter *)cascade;
1447 }
1448
1449 /*
1450 * ident filter
1451 */
1452 #define IDENT_DRAINING (-1)
1453 #define IDENT_SKIPPING (-2)
1454 struct ident_filter {
1455 struct stream_filter filter;
1456 struct strbuf left;
1457 int state;
1458 char ident[45]; /* ": x40 $" */
1459 };
1460
1461 static int is_foreign_ident(const char *str)
1462 {
1463 int i;
1464
1465 if (!skip_prefix(str, "$Id: ", &str))
1466 return 0;
1467 for (i = 0; str[i]; i++) {
1468 if (isspace(str[i]) && str[i+1] != '$')
1469 return 1;
1470 }
1471 return 0;
1472 }
1473
1474 static void ident_drain(struct ident_filter *ident, char **output_p, size_t *osize_p)
1475 {
1476 size_t to_drain = ident->left.len;
1477
1478 if (*osize_p < to_drain)
1479 to_drain = *osize_p;
1480 if (to_drain) {
1481 memcpy(*output_p, ident->left.buf, to_drain);
1482 strbuf_remove(&ident->left, 0, to_drain);
1483 *output_p += to_drain;
1484 *osize_p -= to_drain;
1485 }
1486 if (!ident->left.len)
1487 ident->state = 0;
1488 }
1489
1490 static int ident_filter_fn(struct stream_filter *filter,
1491 const char *input, size_t *isize_p,
1492 char *output, size_t *osize_p)
1493 {
1494 struct ident_filter *ident = (struct ident_filter *)filter;
1495 static const char head[] = "$Id";
1496
1497 if (!input) {
1498 /* drain upon eof */
1499 switch (ident->state) {
1500 default:
1501 strbuf_add(&ident->left, head, ident->state);
1502 case IDENT_SKIPPING:
1503 /* fallthru */
1504 case IDENT_DRAINING:
1505 ident_drain(ident, &output, osize_p);
1506 }
1507 return 0;
1508 }
1509
1510 while (*isize_p || (ident->state == IDENT_DRAINING)) {
1511 int ch;
1512
1513 if (ident->state == IDENT_DRAINING) {
1514 ident_drain(ident, &output, osize_p);
1515 if (!*osize_p)
1516 break;
1517 continue;
1518 }
1519
1520 ch = *(input++);
1521 (*isize_p)--;
1522
1523 if (ident->state == IDENT_SKIPPING) {
1524 /*
1525 * Skipping until '$' or LF, but keeping them
1526 * in case it is a foreign ident.
1527 */
1528 strbuf_addch(&ident->left, ch);
1529 if (ch != '\n' && ch != '$')
1530 continue;
1531 if (ch == '$' && !is_foreign_ident(ident->left.buf)) {
1532 strbuf_setlen(&ident->left, sizeof(head) - 1);
1533 strbuf_addstr(&ident->left, ident->ident);
1534 }
1535 ident->state = IDENT_DRAINING;
1536 continue;
1537 }
1538
1539 if (ident->state < sizeof(head) &&
1540 head[ident->state] == ch) {
1541 ident->state++;
1542 continue;
1543 }
1544
1545 if (ident->state)
1546 strbuf_add(&ident->left, head, ident->state);
1547 if (ident->state == sizeof(head) - 1) {
1548 if (ch != ':' && ch != '$') {
1549 strbuf_addch(&ident->left, ch);
1550 ident->state = 0;
1551 continue;
1552 }
1553
1554 if (ch == ':') {
1555 strbuf_addch(&ident->left, ch);
1556 ident->state = IDENT_SKIPPING;
1557 } else {
1558 strbuf_addstr(&ident->left, ident->ident);
1559 ident->state = IDENT_DRAINING;
1560 }
1561 continue;
1562 }
1563
1564 strbuf_addch(&ident->left, ch);
1565 ident->state = IDENT_DRAINING;
1566 }
1567 return 0;
1568 }
1569
1570 static void ident_free_fn(struct stream_filter *filter)
1571 {
1572 struct ident_filter *ident = (struct ident_filter *)filter;
1573 strbuf_release(&ident->left);
1574 free(filter);
1575 }
1576
1577 static struct stream_filter_vtbl ident_vtbl = {
1578 ident_filter_fn,
1579 ident_free_fn,
1580 };
1581
1582 static struct stream_filter *ident_filter(const unsigned char *sha1)
1583 {
1584 struct ident_filter *ident = xmalloc(sizeof(*ident));
1585
1586 xsnprintf(ident->ident, sizeof(ident->ident),
1587 ": %s $", sha1_to_hex(sha1));
1588 strbuf_init(&ident->left, 0);
1589 ident->filter.vtbl = &ident_vtbl;
1590 ident->state = 0;
1591 return (struct stream_filter *)ident;
1592 }
1593
1594 /*
1595 * Return an appropriately constructed filter for the path, or NULL if
1596 * the contents cannot be filtered without reading the whole thing
1597 * in-core.
1598 *
1599 * Note that you would be crazy to set CRLF, smuge/clean or ident to a
1600 * large binary blob you would want us not to slurp into the memory!
1601 */
1602 struct stream_filter *get_stream_filter(const char *path, const unsigned char *sha1)
1603 {
1604 struct conv_attrs ca;
1605 struct stream_filter *filter = NULL;
1606
1607 convert_attrs(&ca, path);
1608 if (ca.drv && (ca.drv->process || ca.drv->smudge || ca.drv->clean))
1609 return NULL;
1610
1611 if (ca.crlf_action == CRLF_AUTO || ca.crlf_action == CRLF_AUTO_CRLF)
1612 return NULL;
1613
1614 if (ca.ident)
1615 filter = ident_filter(sha1);
1616
1617 if (output_eol(ca.crlf_action) == EOL_CRLF)
1618 filter = cascade_filter(filter, lf_to_crlf_filter());
1619 else
1620 filter = cascade_filter(filter, &null_filter_singleton);
1621
1622 return filter;
1623 }
1624
1625 void free_stream_filter(struct stream_filter *filter)
1626 {
1627 filter->vtbl->free(filter);
1628 }
1629
1630 int stream_filter(struct stream_filter *filter,
1631 const char *input, size_t *isize_p,
1632 char *output, size_t *osize_p)
1633 {
1634 return filter->vtbl->filter(filter, input, isize_p, output, osize_p);
1635 }
git for Debian