1 #define NO_THE_INDEX_COMPATIBILITY_MACROS
5 #include "run-command.h"
9 #include "sub-process.h"
12 * convert.c - convert a file when checking it out and checking it in.
14 * This should use the pathname to decide on whether it wants to do some
15 * more interesting conversions (automatic gzip/unzip, general format
16 * conversions etc etc), but by default it just does automatic CRLF<->LF
17 * translation when the "text" attribute or "auto_crlf" option is set.
20 /* Stat bits: When BIN is set, the txt bits are unset */
21 #define CONVERT_STAT_BITS_TXT_LF 0x1
22 #define CONVERT_STAT_BITS_TXT_CRLF 0x2
23 #define CONVERT_STAT_BITS_BIN 0x4
37 /* NUL, CR, LF and CRLF counts */
38 unsigned nul, lonecr, lonelf, crlf;
40 /* These are just approximations! */
41 unsigned printable, nonprintable;
44 static void gather_stats(const char *buf, unsigned long size, struct text_stat *stats)
48 memset(stats, 0, sizeof(*stats));
50 for (i = 0; i < size; i++) {
51 unsigned char c = buf[i];
53 if (i+1 < size && buf[i+1] == '\n') {
66 stats->nonprintable++;
69 /* BS, HT, ESC and FF */
70 case '\b': case '\t': case '\033': case '\014':
77 stats->nonprintable++;
84 /* If file ends with EOF then don't count this EOF as non-printable. */
85 if (size >= 1 && buf[size-1] == '\032')
86 stats->nonprintable--;
90 * The same heuristics as diff.c::mmfile_is_binary()
91 * We treat files with bare CR as binary
93 static int convert_is_binary(unsigned long size, const struct text_stat *stats)
99 if ((stats->printable >> 7) < stats->nonprintable)
104 static unsigned int gather_convert_stats(const char *data, unsigned long size)
106 struct text_stat stats;
110 gather_stats(data, size, &stats);
111 if (convert_is_binary(size, &stats))
112 ret |= CONVERT_STAT_BITS_BIN;
114 ret |= CONVERT_STAT_BITS_TXT_CRLF;
116 ret |= CONVERT_STAT_BITS_TXT_LF;
121 static const char *gather_convert_stats_ascii(const char *data, unsigned long size)
123 unsigned int convert_stats = gather_convert_stats(data, size);
125 if (convert_stats & CONVERT_STAT_BITS_BIN)
127 switch (convert_stats) {
128 case CONVERT_STAT_BITS_TXT_LF:
130 case CONVERT_STAT_BITS_TXT_CRLF:
132 case CONVERT_STAT_BITS_TXT_LF | CONVERT_STAT_BITS_TXT_CRLF:
139 const char *get_cached_convert_stats_ascii(const struct index_state *istate,
144 void *data = read_blob_data_from_index(istate, path, &sz);
145 ret = gather_convert_stats_ascii(data, sz);
150 const char *get_wt_convert_stats_ascii(const char *path)
152 const char *ret = "";
153 struct strbuf sb = STRBUF_INIT;
154 if (strbuf_read_file(&sb, path, 0) >= 0)
155 ret = gather_convert_stats_ascii(sb.buf, sb.len);
160 static int text_eol_is_crlf(void)
162 if (auto_crlf == AUTO_CRLF_TRUE)
164 else if (auto_crlf == AUTO_CRLF_INPUT)
166 if (core_eol == EOL_CRLF)
168 if (core_eol == EOL_UNSET && EOL_NATIVE == EOL_CRLF)
173 static enum eol output_eol(enum crlf_action crlf_action)
175 switch (crlf_action) {
180 case CRLF_TEXT_INPUT:
185 case CRLF_AUTO_INPUT:
190 return text_eol_is_crlf() ? EOL_CRLF : EOL_LF;
192 warning("Illegal crlf_action %d\n", (int)crlf_action);
196 static void check_safe_crlf(const char *path, enum crlf_action crlf_action,
197 struct text_stat *old_stats, struct text_stat *new_stats,
198 enum safe_crlf checksafe)
200 if (old_stats->crlf && !new_stats->crlf ) {
202 * CRLFs would not be restored by checkout
204 if (checksafe == SAFE_CRLF_WARN)
205 warning(_("CRLF will be replaced by LF in %s.\n"
206 "The file will have its original line"
207 " endings in your working directory."), path);
208 else /* i.e. SAFE_CRLF_FAIL */
209 die(_("CRLF would be replaced by LF in %s."), path);
210 } else if (old_stats->lonelf && !new_stats->lonelf ) {
212 * CRLFs would be added by checkout
214 if (checksafe == SAFE_CRLF_WARN)
215 warning(_("LF will be replaced by CRLF in %s.\n"
216 "The file will have its original line"
217 " endings in your working directory."), path);
218 else /* i.e. SAFE_CRLF_FAIL */
219 die(_("LF would be replaced by CRLF in %s"), path);
223 static int has_cr_in_index(const struct index_state *istate, const char *path)
229 data = read_blob_data_from_index(istate, path, &sz);
232 has_cr = memchr(data, '\r', sz) != NULL;
237 static int will_convert_lf_to_crlf(size_t len, struct text_stat *stats,
238 enum crlf_action crlf_action)
240 if (output_eol(crlf_action) != EOL_CRLF)
242 /* No "naked" LF? Nothing to convert, regardless. */
246 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
247 /* If we have any CR or CRLF line endings, we do not touch it */
248 /* This is the new safer autocrlf-handling */
249 if (stats->lonecr || stats->crlf)
252 if (convert_is_binary(len, stats))
259 static int crlf_to_git(const struct index_state *istate,
260 const char *path, const char *src, size_t len,
262 enum crlf_action crlf_action, enum safe_crlf checksafe)
264 struct text_stat stats;
266 int convert_crlf_into_lf;
268 if (crlf_action == CRLF_BINARY ||
273 * If we are doing a dry-run and have no source buffer, there is
274 * nothing to analyze; we must assume we would convert.
279 gather_stats(src, len, &stats);
280 /* Optimization: No CRLF? Nothing to convert, regardless. */
281 convert_crlf_into_lf = !!stats.crlf;
283 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
284 if (convert_is_binary(len, &stats))
287 * If the file in the index has any CR in it, do not
288 * convert. This is the new safer autocrlf handling,
289 * unless we want to renormalize in a merge or
292 if ((checksafe != SAFE_CRLF_RENORMALIZE) &&
293 has_cr_in_index(istate, path))
294 convert_crlf_into_lf = 0;
296 if ((checksafe == SAFE_CRLF_WARN ||
297 (checksafe == SAFE_CRLF_FAIL)) && len) {
298 struct text_stat new_stats;
299 memcpy(&new_stats, &stats, sizeof(new_stats));
300 /* simulate "git add" */
301 if (convert_crlf_into_lf) {
302 new_stats.lonelf += new_stats.crlf;
305 /* simulate "git checkout" */
306 if (will_convert_lf_to_crlf(len, &new_stats, crlf_action)) {
307 new_stats.crlf += new_stats.lonelf;
308 new_stats.lonelf = 0;
310 check_safe_crlf(path, crlf_action, &stats, &new_stats, checksafe);
312 if (!convert_crlf_into_lf)
316 * At this point all of our source analysis is done, and we are sure we
317 * would convert. If we are in dry-run mode, we can give an answer.
322 /* only grow if not in place */
323 if (strbuf_avail(buf) + buf->len < len)
324 strbuf_grow(buf, len - buf->len);
326 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
328 * If we guessed, we already know we rejected a file with
329 * lone CR, and we can strip a CR without looking at what
333 unsigned char c = *src++;
339 unsigned char c = *src++;
340 if (! (c == '\r' && (1 < len && *src == '\n')))
344 strbuf_setlen(buf, dst - buf->buf);
348 static int crlf_to_worktree(const char *path, const char *src, size_t len,
349 struct strbuf *buf, enum crlf_action crlf_action)
351 char *to_free = NULL;
352 struct text_stat stats;
354 if (!len || output_eol(crlf_action) != EOL_CRLF)
357 gather_stats(src, len, &stats);
358 if (!will_convert_lf_to_crlf(len, &stats, crlf_action))
361 /* are we "faking" in place editing ? */
363 to_free = strbuf_detach(buf, NULL);
365 strbuf_grow(buf, len + stats.lonelf);
367 const char *nl = memchr(src, '\n', len);
370 if (nl > src && nl[-1] == '\r') {
371 strbuf_add(buf, src, nl + 1 - src);
373 strbuf_add(buf, src, nl - src);
374 strbuf_addstr(buf, "\r\n");
379 strbuf_add(buf, src, len);
385 struct filter_params {
393 static int filter_buffer_or_fd(int in, int out, void *data)
396 * Spawn cmd and feed the buffer contents through its stdin.
398 struct child_process child_process = CHILD_PROCESS_INIT;
399 struct filter_params *params = (struct filter_params *)data;
400 int write_err, status;
401 const char *argv[] = { NULL, NULL };
403 /* apply % substitution to cmd */
404 struct strbuf cmd = STRBUF_INIT;
405 struct strbuf path = STRBUF_INIT;
406 struct strbuf_expand_dict_entry dict[] = {
411 /* quote the path to preserve spaces, etc. */
412 sq_quote_buf(&path, params->path);
413 dict[0].value = path.buf;
415 /* expand all %f with the quoted path */
416 strbuf_expand(&cmd, params->cmd, strbuf_expand_dict_cb, &dict);
417 strbuf_release(&path);
421 child_process.argv = argv;
422 child_process.use_shell = 1;
423 child_process.in = -1;
424 child_process.out = out;
426 if (start_command(&child_process)) {
427 strbuf_release(&cmd);
428 return error("cannot fork to run external filter '%s'", params->cmd);
431 sigchain_push(SIGPIPE, SIG_IGN);
434 write_err = (write_in_full(child_process.in,
435 params->src, params->size) < 0);
439 write_err = copy_fd(params->fd, child_process.in);
440 if (write_err == COPY_WRITE_ERROR && errno == EPIPE)
444 if (close(child_process.in))
447 error("cannot feed the input to external filter '%s'", params->cmd);
449 sigchain_pop(SIGPIPE);
451 status = finish_command(&child_process);
453 error("external filter '%s' failed %d", params->cmd, status);
455 strbuf_release(&cmd);
456 return (write_err || status);
459 static int apply_single_file_filter(const char *path, const char *src, size_t len, int fd,
460 struct strbuf *dst, const char *cmd)
463 * Create a pipeline to have the command filter the buffer's
466 * (child --> cmd) --> us
469 struct strbuf nbuf = STRBUF_INIT;
471 struct filter_params params;
473 memset(&async, 0, sizeof(async));
474 async.proc = filter_buffer_or_fd;
475 async.data = ¶ms;
484 if (start_async(&async))
485 return 0; /* error was already reported */
487 if (strbuf_read(&nbuf, async.out, len) < 0) {
488 err = error("read from external filter '%s' failed", cmd);
490 if (close(async.out)) {
491 err = error("read from external filter '%s' failed", cmd);
493 if (finish_async(&async)) {
494 err = error("external filter '%s' failed", cmd);
498 strbuf_swap(dst, &nbuf);
500 strbuf_release(&nbuf);
504 #define CAP_CLEAN (1u<<0)
505 #define CAP_SMUDGE (1u<<1)
506 #define CAP_DELAY (1u<<2)
509 struct subprocess_entry subprocess; /* must be the first member! */
510 unsigned int supported_capabilities;
513 static int subprocess_map_initialized;
514 static struct hashmap subprocess_map;
516 static int start_multi_file_filter_fn(struct subprocess_entry *subprocess)
518 static int versions[] = {2, 0};
519 static struct subprocess_capability capabilities[] = {
520 { "clean", CAP_CLEAN },
521 { "smudge", CAP_SMUDGE },
522 { "delay", CAP_DELAY },
525 struct cmd2process *entry = (struct cmd2process *)subprocess;
526 return subprocess_handshake(subprocess, "git-filter", versions, NULL,
528 &entry->supported_capabilities);
531 static void handle_filter_error(const struct strbuf *filter_status,
532 struct cmd2process *entry,
533 const unsigned int wanted_capability) {
534 if (!strcmp(filter_status->buf, "error"))
535 ; /* The filter signaled a problem with the file. */
536 else if (!strcmp(filter_status->buf, "abort") && wanted_capability) {
538 * The filter signaled a permanent problem. Don't try to filter
539 * files with the same command for the lifetime of the current
542 entry->supported_capabilities &= ~wanted_capability;
545 * Something went wrong with the protocol filter.
546 * Force shutdown and restart if another blob requires filtering.
548 error("external filter '%s' failed", entry->subprocess.cmd);
549 subprocess_stop(&subprocess_map, &entry->subprocess);
554 static int apply_multi_file_filter(const char *path, const char *src, size_t len,
555 int fd, struct strbuf *dst, const char *cmd,
556 const unsigned int wanted_capability,
557 struct delayed_checkout *dco)
561 struct cmd2process *entry;
562 struct child_process *process;
563 struct strbuf nbuf = STRBUF_INIT;
564 struct strbuf filter_status = STRBUF_INIT;
565 const char *filter_type;
567 if (!subprocess_map_initialized) {
568 subprocess_map_initialized = 1;
569 hashmap_init(&subprocess_map, cmd2process_cmp, NULL, 0);
572 entry = (struct cmd2process *)subprocess_find_entry(&subprocess_map, cmd);
578 entry = xmalloc(sizeof(*entry));
579 entry->supported_capabilities = 0;
581 if (subprocess_start(&subprocess_map, &entry->subprocess, cmd, start_multi_file_filter_fn)) {
586 process = &entry->subprocess.process;
588 if (!(entry->supported_capabilities & wanted_capability))
591 if (wanted_capability & CAP_CLEAN)
592 filter_type = "clean";
593 else if (wanted_capability & CAP_SMUDGE)
594 filter_type = "smudge";
596 die("unexpected filter type");
598 sigchain_push(SIGPIPE, SIG_IGN);
600 assert(strlen(filter_type) < LARGE_PACKET_DATA_MAX - strlen("command=\n"));
601 err = packet_write_fmt_gently(process->in, "command=%s\n", filter_type);
605 err = strlen(path) > LARGE_PACKET_DATA_MAX - strlen("pathname=\n");
607 error("path name too long for external filter");
611 err = packet_write_fmt_gently(process->in, "pathname=%s\n", path);
615 if ((entry->supported_capabilities & CAP_DELAY) &&
616 dco && dco->state == CE_CAN_DELAY) {
618 err = packet_write_fmt_gently(process->in, "can-delay=1\n");
623 err = packet_flush_gently(process->in);
628 err = write_packetized_from_fd(fd, process->in);
630 err = write_packetized_from_buf(src, len, process->in);
634 err = subprocess_read_status(process->out, &filter_status);
638 if (can_delay && !strcmp(filter_status.buf, "delayed")) {
639 string_list_insert(&dco->filters, cmd);
640 string_list_insert(&dco->paths, path);
642 /* The filter got the blob and wants to send us a response. */
643 err = strcmp(filter_status.buf, "success");
647 err = read_packetized_to_strbuf(process->out, &nbuf) < 0;
651 err = subprocess_read_status(process->out, &filter_status);
655 err = strcmp(filter_status.buf, "success");
659 sigchain_pop(SIGPIPE);
662 handle_filter_error(&filter_status, entry, wanted_capability);
664 strbuf_swap(dst, &nbuf);
665 strbuf_release(&nbuf);
670 int async_query_available_blobs(const char *cmd, struct string_list *available_paths)
674 struct cmd2process *entry;
675 struct child_process *process;
676 struct strbuf filter_status = STRBUF_INIT;
678 assert(subprocess_map_initialized);
679 entry = (struct cmd2process *)subprocess_find_entry(&subprocess_map, cmd);
681 error("external filter '%s' is not available anymore although "
682 "not all paths have been filtered", cmd);
685 process = &entry->subprocess.process;
686 sigchain_push(SIGPIPE, SIG_IGN);
688 err = packet_write_fmt_gently(
689 process->in, "command=list_available_blobs\n");
693 err = packet_flush_gently(process->in);
697 while ((line = packet_read_line(process->out, NULL))) {
699 if (skip_prefix(line, "pathname=", &path))
700 string_list_insert(available_paths, xstrdup(path));
702 ; /* ignore unknown keys */
705 err = subprocess_read_status(process->out, &filter_status);
709 err = strcmp(filter_status.buf, "success");
712 sigchain_pop(SIGPIPE);
715 handle_filter_error(&filter_status, entry, 0);
719 static struct convert_driver {
721 struct convert_driver *next;
726 } *user_convert, **user_convert_tail;
728 static int apply_filter(const char *path, const char *src, size_t len,
729 int fd, struct strbuf *dst, struct convert_driver *drv,
730 const unsigned int wanted_capability,
731 struct delayed_checkout *dco)
733 const char *cmd = NULL;
741 if ((wanted_capability & CAP_CLEAN) && !drv->process && drv->clean)
743 else if ((wanted_capability & CAP_SMUDGE) && !drv->process && drv->smudge)
747 return apply_single_file_filter(path, src, len, fd, dst, cmd);
748 else if (drv->process && *drv->process)
749 return apply_multi_file_filter(path, src, len, fd, dst,
750 drv->process, wanted_capability, dco);
755 static int read_convert_config(const char *var, const char *value, void *cb)
757 const char *key, *name;
759 struct convert_driver *drv;
762 * External conversion drivers are configured using
763 * "filter.<name>.variable".
765 if (parse_config_key(var, "filter", &name, &namelen, &key) < 0 || !name)
767 for (drv = user_convert; drv; drv = drv->next)
768 if (!strncmp(drv->name, name, namelen) && !drv->name[namelen])
771 drv = xcalloc(1, sizeof(struct convert_driver));
772 drv->name = xmemdupz(name, namelen);
773 *user_convert_tail = drv;
774 user_convert_tail = &(drv->next);
778 * filter.<name>.smudge and filter.<name>.clean specifies
783 * The command-line will not be interpolated in any way.
786 if (!strcmp("smudge", key))
787 return git_config_string(&drv->smudge, var, value);
789 if (!strcmp("clean", key))
790 return git_config_string(&drv->clean, var, value);
792 if (!strcmp("process", key))
793 return git_config_string(&drv->process, var, value);
795 if (!strcmp("required", key)) {
796 drv->required = git_config_bool(var, value);
803 static int count_ident(const char *cp, unsigned long size)
806 * "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$"
818 if (memcmp("Id", cp, 2))
829 * "$Id: ... "; scan up to the closing dollar sign and discard.
845 static int ident_to_git(const char *path, const char *src, size_t len,
846 struct strbuf *buf, int ident)
850 if (!ident || (src && !count_ident(src, len)))
856 /* only grow if not in place */
857 if (strbuf_avail(buf) + buf->len < len)
858 strbuf_grow(buf, len - buf->len);
861 dollar = memchr(src, '$', len);
864 memmove(dst, src, dollar + 1 - src);
865 dst += dollar + 1 - src;
866 len -= dollar + 1 - src;
869 if (len > 3 && !memcmp(src, "Id:", 3)) {
870 dollar = memchr(src + 3, '$', len - 3);
873 if (memchr(src + 3, '\n', dollar - src - 3)) {
874 /* Line break before the next dollar. */
878 memcpy(dst, "Id$", 3);
880 len -= dollar + 1 - src;
884 memmove(dst, src, len);
885 strbuf_setlen(buf, dst + len - buf->buf);
889 static int ident_to_worktree(const char *path, const char *src, size_t len,
890 struct strbuf *buf, int ident)
892 unsigned char sha1[20];
893 char *to_free = NULL, *dollar, *spc;
899 cnt = count_ident(src, len);
903 /* are we "faking" in place editing ? */
905 to_free = strbuf_detach(buf, NULL);
906 hash_sha1_file(src, len, "blob", sha1);
908 strbuf_grow(buf, len + cnt * 43);
910 /* step 1: run to the next '$' */
911 dollar = memchr(src, '$', len);
914 strbuf_add(buf, src, dollar + 1 - src);
915 len -= dollar + 1 - src;
918 /* step 2: does it looks like a bit like Id:xxx$ or Id$ ? */
919 if (len < 3 || memcmp("Id", src, 2))
922 /* step 3: skip over Id$ or Id:xxxxx$ */
926 } else if (src[2] == ':') {
928 * It's possible that an expanded Id has crept its way into the
929 * repository, we cope with that by stripping the expansion out.
930 * This is probably not a good idea, since it will cause changes
931 * on checkout, which won't go away by stash, but let's keep it
934 dollar = memchr(src + 3, '$', len - 3);
936 /* incomplete keyword, no more '$', so just quit the loop */
940 if (memchr(src + 3, '\n', dollar - src - 3)) {
941 /* Line break before the next dollar. */
945 spc = memchr(src + 4, ' ', dollar - src - 4);
946 if (spc && spc < dollar-1) {
947 /* There are spaces in unexpected places.
948 * This is probably an id from some other
949 * versioning system. Keep it for now.
954 len -= dollar + 1 - src;
957 /* it wasn't a "Id$" or "Id:xxxx$" */
961 /* step 4: substitute */
962 strbuf_addstr(buf, "Id: ");
963 strbuf_add(buf, sha1_to_hex(sha1), 40);
964 strbuf_addstr(buf, " $");
966 strbuf_add(buf, src, len);
972 static enum crlf_action git_path_check_crlf(struct attr_check_item *check)
974 const char *value = check->value;
976 if (ATTR_TRUE(value))
978 else if (ATTR_FALSE(value))
980 else if (ATTR_UNSET(value))
982 else if (!strcmp(value, "input"))
983 return CRLF_TEXT_INPUT;
984 else if (!strcmp(value, "auto"))
986 return CRLF_UNDEFINED;
989 static enum eol git_path_check_eol(struct attr_check_item *check)
991 const char *value = check->value;
993 if (ATTR_UNSET(value))
995 else if (!strcmp(value, "lf"))
997 else if (!strcmp(value, "crlf"))
1002 static struct convert_driver *git_path_check_convert(struct attr_check_item *check)
1004 const char *value = check->value;
1005 struct convert_driver *drv;
1007 if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
1009 for (drv = user_convert; drv; drv = drv->next)
1010 if (!strcmp(value, drv->name))
1015 static int git_path_check_ident(struct attr_check_item *check)
1017 const char *value = check->value;
1019 return !!ATTR_TRUE(value);
1023 struct convert_driver *drv;
1024 enum crlf_action attr_action; /* What attr says */
1025 enum crlf_action crlf_action; /* When no attr is set, use core.autocrlf */
1029 static void convert_attrs(struct conv_attrs *ca, const char *path)
1031 static struct attr_check *check;
1034 check = attr_check_initl("crlf", "ident", "filter",
1035 "eol", "text", NULL);
1036 user_convert_tail = &user_convert;
1037 git_config(read_convert_config, NULL);
1040 if (!git_check_attr(path, check)) {
1041 struct attr_check_item *ccheck = check->items;
1042 ca->crlf_action = git_path_check_crlf(ccheck + 4);
1043 if (ca->crlf_action == CRLF_UNDEFINED)
1044 ca->crlf_action = git_path_check_crlf(ccheck + 0);
1045 ca->ident = git_path_check_ident(ccheck + 1);
1046 ca->drv = git_path_check_convert(ccheck + 2);
1047 if (ca->crlf_action != CRLF_BINARY) {
1048 enum eol eol_attr = git_path_check_eol(ccheck + 3);
1049 if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_LF)
1050 ca->crlf_action = CRLF_AUTO_INPUT;
1051 else if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_CRLF)
1052 ca->crlf_action = CRLF_AUTO_CRLF;
1053 else if (eol_attr == EOL_LF)
1054 ca->crlf_action = CRLF_TEXT_INPUT;
1055 else if (eol_attr == EOL_CRLF)
1056 ca->crlf_action = CRLF_TEXT_CRLF;
1060 ca->crlf_action = CRLF_UNDEFINED;
1064 /* Save attr and make a decision for action */
1065 ca->attr_action = ca->crlf_action;
1066 if (ca->crlf_action == CRLF_TEXT)
1067 ca->crlf_action = text_eol_is_crlf() ? CRLF_TEXT_CRLF : CRLF_TEXT_INPUT;
1068 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_FALSE)
1069 ca->crlf_action = CRLF_BINARY;
1070 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_TRUE)
1071 ca->crlf_action = CRLF_AUTO_CRLF;
1072 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_INPUT)
1073 ca->crlf_action = CRLF_AUTO_INPUT;
1076 int would_convert_to_git_filter_fd(const char *path)
1078 struct conv_attrs ca;
1080 convert_attrs(&ca, path);
1085 * Apply a filter to an fd only if the filter is required to succeed.
1086 * We must die if the filter fails, because the original data before
1087 * filtering is not available.
1089 if (!ca.drv->required)
1092 return apply_filter(path, NULL, 0, -1, NULL, ca.drv, CAP_CLEAN, NULL);
1095 const char *get_convert_attr_ascii(const char *path)
1097 struct conv_attrs ca;
1099 convert_attrs(&ca, path);
1100 switch (ca.attr_action) {
1101 case CRLF_UNDEFINED:
1107 case CRLF_TEXT_INPUT:
1108 return "text eol=lf";
1109 case CRLF_TEXT_CRLF:
1110 return "text eol=crlf";
1113 case CRLF_AUTO_CRLF:
1114 return "text=auto eol=crlf";
1115 case CRLF_AUTO_INPUT:
1116 return "text=auto eol=lf";
1121 int convert_to_git(const struct index_state *istate,
1122 const char *path, const char *src, size_t len,
1123 struct strbuf *dst, enum safe_crlf checksafe)
1126 struct conv_attrs ca;
1128 convert_attrs(&ca, path);
1130 ret |= apply_filter(path, src, len, -1, dst, ca.drv, CAP_CLEAN, NULL);
1131 if (!ret && ca.drv && ca.drv->required)
1132 die("%s: clean filter '%s' failed", path, ca.drv->name);
1138 if (checksafe != SAFE_CRLF_KEEP_CRLF) {
1139 ret |= crlf_to_git(istate, path, src, len, dst, ca.crlf_action, checksafe);
1145 return ret | ident_to_git(path, src, len, dst, ca.ident);
1148 void convert_to_git_filter_fd(const struct index_state *istate,
1149 const char *path, int fd, struct strbuf *dst,
1150 enum safe_crlf checksafe)
1152 struct conv_attrs ca;
1153 convert_attrs(&ca, path);
1156 assert(ca.drv->clean || ca.drv->process);
1158 if (!apply_filter(path, NULL, 0, fd, dst, ca.drv, CAP_CLEAN, NULL))
1159 die("%s: clean filter '%s' failed", path, ca.drv->name);
1161 crlf_to_git(istate, path, dst->buf, dst->len, dst, ca.crlf_action, checksafe);
1162 ident_to_git(path, dst->buf, dst->len, dst, ca.ident);
1165 static int convert_to_working_tree_internal(const char *path, const char *src,
1166 size_t len, struct strbuf *dst,
1167 int normalizing, struct delayed_checkout *dco)
1169 int ret = 0, ret_filter = 0;
1170 struct conv_attrs ca;
1172 convert_attrs(&ca, path);
1174 ret |= ident_to_worktree(path, src, len, dst, ca.ident);
1180 * CRLF conversion can be skipped if normalizing, unless there
1181 * is a smudge or process filter (even if the process filter doesn't
1182 * support smudge). The filters might expect CRLFs.
1184 if ((ca.drv && (ca.drv->smudge || ca.drv->process)) || !normalizing) {
1185 ret |= crlf_to_worktree(path, src, len, dst, ca.crlf_action);
1192 ret_filter = apply_filter(
1193 path, src, len, -1, dst, ca.drv, CAP_SMUDGE, dco);
1194 if (!ret_filter && ca.drv && ca.drv->required)
1195 die("%s: smudge filter %s failed", path, ca.drv->name);
1197 return ret | ret_filter;
1200 int async_convert_to_working_tree(const char *path, const char *src,
1201 size_t len, struct strbuf *dst,
1204 return convert_to_working_tree_internal(path, src, len, dst, 0, dco);
1207 int convert_to_working_tree(const char *path, const char *src, size_t len, struct strbuf *dst)
1209 return convert_to_working_tree_internal(path, src, len, dst, 0, NULL);
1212 int renormalize_buffer(const struct index_state *istate, const char *path,
1213 const char *src, size_t len, struct strbuf *dst)
1215 int ret = convert_to_working_tree_internal(path, src, len, dst, 1, NULL);
1220 return ret | convert_to_git(istate, path, src, len, dst, SAFE_CRLF_RENORMALIZE);
1223 /*****************************************************************
1225 * Streaming conversion support
1227 *****************************************************************/
1229 typedef int (*filter_fn)(struct stream_filter *,
1230 const char *input, size_t *isize_p,
1231 char *output, size_t *osize_p);
1232 typedef void (*free_fn)(struct stream_filter *);
1234 struct stream_filter_vtbl {
1239 struct stream_filter {
1240 struct stream_filter_vtbl *vtbl;
1243 static int null_filter_fn(struct stream_filter *filter,
1244 const char *input, size_t *isize_p,
1245 char *output, size_t *osize_p)
1250 return 0; /* we do not keep any states */
1252 if (*osize_p < count)
1255 memmove(output, input, count);
1262 static void null_free_fn(struct stream_filter *filter)
1264 ; /* nothing -- null instances are shared */
1267 static struct stream_filter_vtbl null_vtbl = {
1272 static struct stream_filter null_filter_singleton = {
1276 int is_null_stream_filter(struct stream_filter *filter)
1278 return filter == &null_filter_singleton;
1286 struct lf_to_crlf_filter {
1287 struct stream_filter filter;
1288 unsigned has_held:1;
1292 static int lf_to_crlf_filter_fn(struct stream_filter *filter,
1293 const char *input, size_t *isize_p,
1294 char *output, size_t *osize_p)
1296 size_t count, o = 0;
1297 struct lf_to_crlf_filter *lf_to_crlf = (struct lf_to_crlf_filter *)filter;
1300 * We may be holding onto the CR to see if it is followed by a
1301 * LF, in which case we would need to go to the main loop.
1302 * Otherwise, just emit it to the output stream.
1304 if (lf_to_crlf->has_held && (lf_to_crlf->held != '\r' || !input)) {
1305 output[o++] = lf_to_crlf->held;
1306 lf_to_crlf->has_held = 0;
1309 /* We are told to drain */
1316 if (count || lf_to_crlf->has_held) {
1320 if (lf_to_crlf->has_held) {
1322 lf_to_crlf->has_held = 0;
1325 for (i = 0; o < *osize_p && i < count; i++) {
1330 } else if (was_cr) {
1332 * Previous round saw CR and it is not followed
1333 * by a LF; emit the CR before processing the
1334 * current character.
1340 * We may have consumed the last output slot,
1341 * in which case we need to break out of this
1342 * loop; hold the current character before
1345 if (*osize_p <= o) {
1346 lf_to_crlf->has_held = 1;
1347 lf_to_crlf->held = ch;
1348 continue; /* break but increment i */
1363 if (!lf_to_crlf->has_held && was_cr) {
1364 lf_to_crlf->has_held = 1;
1365 lf_to_crlf->held = '\r';
1371 static void lf_to_crlf_free_fn(struct stream_filter *filter)
1376 static struct stream_filter_vtbl lf_to_crlf_vtbl = {
1377 lf_to_crlf_filter_fn,
1381 static struct stream_filter *lf_to_crlf_filter(void)
1383 struct lf_to_crlf_filter *lf_to_crlf = xcalloc(1, sizeof(*lf_to_crlf));
1385 lf_to_crlf->filter.vtbl = &lf_to_crlf_vtbl;
1386 return (struct stream_filter *)lf_to_crlf;
1392 #define FILTER_BUFFER 1024
1393 struct cascade_filter {
1394 struct stream_filter filter;
1395 struct stream_filter *one;
1396 struct stream_filter *two;
1397 char buf[FILTER_BUFFER];
1401 static int cascade_filter_fn(struct stream_filter *filter,
1402 const char *input, size_t *isize_p,
1403 char *output, size_t *osize_p)
1405 struct cascade_filter *cas = (struct cascade_filter *) filter;
1407 size_t sz = *osize_p;
1408 size_t to_feed, remaining;
1411 * input -- (one) --> buf -- (two) --> output
1413 while (filled < sz) {
1414 remaining = sz - filled;
1416 /* do we already have something to feed two with? */
1417 if (cas->ptr < cas->end) {
1418 to_feed = cas->end - cas->ptr;
1419 if (stream_filter(cas->two,
1420 cas->buf + cas->ptr, &to_feed,
1421 output + filled, &remaining))
1423 cas->ptr += (cas->end - cas->ptr) - to_feed;
1424 filled = sz - remaining;
1428 /* feed one from upstream and have it emit into our buffer */
1429 to_feed = input ? *isize_p : 0;
1430 if (input && !to_feed)
1432 remaining = sizeof(cas->buf);
1433 if (stream_filter(cas->one,
1435 cas->buf, &remaining))
1437 cas->end = sizeof(cas->buf) - remaining;
1440 size_t fed = *isize_p - to_feed;
1445 /* do we know that we drained one completely? */
1446 if (input || cas->end)
1449 /* tell two to drain; we have nothing more to give it */
1451 remaining = sz - filled;
1452 if (stream_filter(cas->two,
1454 output + filled, &remaining))
1456 if (remaining == (sz - filled))
1457 break; /* completely drained two */
1458 filled = sz - remaining;
1464 static void cascade_free_fn(struct stream_filter *filter)
1466 struct cascade_filter *cas = (struct cascade_filter *)filter;
1467 free_stream_filter(cas->one);
1468 free_stream_filter(cas->two);
1472 static struct stream_filter_vtbl cascade_vtbl = {
1477 static struct stream_filter *cascade_filter(struct stream_filter *one,
1478 struct stream_filter *two)
1480 struct cascade_filter *cascade;
1482 if (!one || is_null_stream_filter(one))
1484 if (!two || is_null_stream_filter(two))
1487 cascade = xmalloc(sizeof(*cascade));
1490 cascade->end = cascade->ptr = 0;
1491 cascade->filter.vtbl = &cascade_vtbl;
1492 return (struct stream_filter *)cascade;
1498 #define IDENT_DRAINING (-1)
1499 #define IDENT_SKIPPING (-2)
1500 struct ident_filter {
1501 struct stream_filter filter;
1504 char ident[45]; /* ": x40 $" */
1507 static int is_foreign_ident(const char *str)
1511 if (!skip_prefix(str, "$Id: ", &str))
1513 for (i = 0; str[i]; i++) {
1514 if (isspace(str[i]) && str[i+1] != '$')
1520 static void ident_drain(struct ident_filter *ident, char **output_p, size_t *osize_p)
1522 size_t to_drain = ident->left.len;
1524 if (*osize_p < to_drain)
1525 to_drain = *osize_p;
1527 memcpy(*output_p, ident->left.buf, to_drain);
1528 strbuf_remove(&ident->left, 0, to_drain);
1529 *output_p += to_drain;
1530 *osize_p -= to_drain;
1532 if (!ident->left.len)
1536 static int ident_filter_fn(struct stream_filter *filter,
1537 const char *input, size_t *isize_p,
1538 char *output, size_t *osize_p)
1540 struct ident_filter *ident = (struct ident_filter *)filter;
1541 static const char head[] = "$Id";
1544 /* drain upon eof */
1545 switch (ident->state) {
1547 strbuf_add(&ident->left, head, ident->state);
1548 case IDENT_SKIPPING:
1550 case IDENT_DRAINING:
1551 ident_drain(ident, &output, osize_p);
1556 while (*isize_p || (ident->state == IDENT_DRAINING)) {
1559 if (ident->state == IDENT_DRAINING) {
1560 ident_drain(ident, &output, osize_p);
1569 if (ident->state == IDENT_SKIPPING) {
1571 * Skipping until '$' or LF, but keeping them
1572 * in case it is a foreign ident.
1574 strbuf_addch(&ident->left, ch);
1575 if (ch != '\n' && ch != '$')
1577 if (ch == '$' && !is_foreign_ident(ident->left.buf)) {
1578 strbuf_setlen(&ident->left, sizeof(head) - 1);
1579 strbuf_addstr(&ident->left, ident->ident);
1581 ident->state = IDENT_DRAINING;
1585 if (ident->state < sizeof(head) &&
1586 head[ident->state] == ch) {
1592 strbuf_add(&ident->left, head, ident->state);
1593 if (ident->state == sizeof(head) - 1) {
1594 if (ch != ':' && ch != '$') {
1595 strbuf_addch(&ident->left, ch);
1601 strbuf_addch(&ident->left, ch);
1602 ident->state = IDENT_SKIPPING;
1604 strbuf_addstr(&ident->left, ident->ident);
1605 ident->state = IDENT_DRAINING;
1610 strbuf_addch(&ident->left, ch);
1611 ident->state = IDENT_DRAINING;
1616 static void ident_free_fn(struct stream_filter *filter)
1618 struct ident_filter *ident = (struct ident_filter *)filter;
1619 strbuf_release(&ident->left);
1623 static struct stream_filter_vtbl ident_vtbl = {
1628 static struct stream_filter *ident_filter(const unsigned char *sha1)
1630 struct ident_filter *ident = xmalloc(sizeof(*ident));
1632 xsnprintf(ident->ident, sizeof(ident->ident),
1633 ": %s $", sha1_to_hex(sha1));
1634 strbuf_init(&ident->left, 0);
1635 ident->filter.vtbl = &ident_vtbl;
1637 return (struct stream_filter *)ident;
1641 * Return an appropriately constructed filter for the path, or NULL if
1642 * the contents cannot be filtered without reading the whole thing
1645 * Note that you would be crazy to set CRLF, smuge/clean or ident to a
1646 * large binary blob you would want us not to slurp into the memory!
1648 struct stream_filter *get_stream_filter(const char *path, const unsigned char *sha1)
1650 struct conv_attrs ca;
1651 struct stream_filter *filter = NULL;
1653 convert_attrs(&ca, path);
1654 if (ca.drv && (ca.drv->process || ca.drv->smudge || ca.drv->clean))
1657 if (ca.crlf_action == CRLF_AUTO || ca.crlf_action == CRLF_AUTO_CRLF)
1661 filter = ident_filter(sha1);
1663 if (output_eol(ca.crlf_action) == EOL_CRLF)
1664 filter = cascade_filter(filter, lf_to_crlf_filter());
1666 filter = cascade_filter(filter, &null_filter_singleton);
1671 void free_stream_filter(struct stream_filter *filter)
1673 filter->vtbl->free(filter);
1676 int stream_filter(struct stream_filter *filter,
1677 const char *input, size_t *isize_p,
1678 char *output, size_t *osize_p)
1680 return filter->vtbl->filter(filter, input, isize_p, output, osize_p);