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 return error("cannot fork to run external filter '%s'", params->cmd);
429 sigchain_push(SIGPIPE, SIG_IGN);
432 write_err = (write_in_full(child_process.in,
433 params->src, params->size) < 0);
437 write_err = copy_fd(params->fd, child_process.in);
438 if (write_err == COPY_WRITE_ERROR && errno == EPIPE)
442 if (close(child_process.in))
445 error("cannot feed the input to external filter '%s'", params->cmd);
447 sigchain_pop(SIGPIPE);
449 status = finish_command(&child_process);
451 error("external filter '%s' failed %d", params->cmd, status);
453 strbuf_release(&cmd);
454 return (write_err || status);
457 static int apply_single_file_filter(const char *path, const char *src, size_t len, int fd,
458 struct strbuf *dst, const char *cmd)
461 * Create a pipeline to have the command filter the buffer's
464 * (child --> cmd) --> us
467 struct strbuf nbuf = STRBUF_INIT;
469 struct filter_params params;
471 memset(&async, 0, sizeof(async));
472 async.proc = filter_buffer_or_fd;
473 async.data = ¶ms;
482 if (start_async(&async))
483 return 0; /* error was already reported */
485 if (strbuf_read(&nbuf, async.out, len) < 0) {
486 err = error("read from external filter '%s' failed", cmd);
488 if (close(async.out)) {
489 err = error("read from external filter '%s' failed", cmd);
491 if (finish_async(&async)) {
492 err = error("external filter '%s' failed", cmd);
496 strbuf_swap(dst, &nbuf);
498 strbuf_release(&nbuf);
502 #define CAP_CLEAN (1u<<0)
503 #define CAP_SMUDGE (1u<<1)
506 struct subprocess_entry subprocess; /* must be the first member! */
507 unsigned int supported_capabilities;
510 static int subprocess_map_initialized;
511 static struct hashmap subprocess_map;
513 static int start_multi_file_filter_fn(struct subprocess_entry *subprocess)
516 struct cmd2process *entry = (struct cmd2process *)subprocess;
517 struct string_list cap_list = STRING_LIST_INIT_NODUP;
519 const char *cap_name;
520 struct child_process *process = &subprocess->process;
521 const char *cmd = subprocess->cmd;
523 sigchain_push(SIGPIPE, SIG_IGN);
525 err = packet_writel(process->in, "git-filter-client", "version=2", NULL);
529 err = strcmp(packet_read_line(process->out, NULL), "git-filter-server");
531 error("external filter '%s' does not support filter protocol version 2", cmd);
534 err = strcmp(packet_read_line(process->out, NULL), "version=2");
537 err = packet_read_line(process->out, NULL) != NULL;
541 err = packet_writel(process->in, "capability=clean", "capability=smudge", NULL);
544 cap_buf = packet_read_line(process->out, NULL);
547 string_list_split_in_place(&cap_list, cap_buf, '=', 1);
549 if (cap_list.nr != 2 || strcmp(cap_list.items[0].string, "capability"))
552 cap_name = cap_list.items[1].string;
553 if (!strcmp(cap_name, "clean")) {
554 entry->supported_capabilities |= CAP_CLEAN;
555 } else if (!strcmp(cap_name, "smudge")) {
556 entry->supported_capabilities |= CAP_SMUDGE;
559 "external filter '%s' requested unsupported filter capability '%s'",
564 string_list_clear(&cap_list, 0);
568 sigchain_pop(SIGPIPE);
573 static int apply_multi_file_filter(const char *path, const char *src, size_t len,
574 int fd, struct strbuf *dst, const char *cmd,
575 const unsigned int wanted_capability)
578 struct cmd2process *entry;
579 struct child_process *process;
580 struct strbuf nbuf = STRBUF_INIT;
581 struct strbuf filter_status = STRBUF_INIT;
582 const char *filter_type;
584 if (!subprocess_map_initialized) {
585 subprocess_map_initialized = 1;
586 hashmap_init(&subprocess_map, (hashmap_cmp_fn) cmd2process_cmp, 0);
589 entry = (struct cmd2process *)subprocess_find_entry(&subprocess_map, cmd);
595 entry = xmalloc(sizeof(*entry));
596 entry->supported_capabilities = 0;
598 if (subprocess_start(&subprocess_map, &entry->subprocess, cmd, start_multi_file_filter_fn)) {
603 process = &entry->subprocess.process;
605 if (!(wanted_capability & entry->supported_capabilities))
608 if (CAP_CLEAN & wanted_capability)
609 filter_type = "clean";
610 else if (CAP_SMUDGE & wanted_capability)
611 filter_type = "smudge";
613 die("unexpected filter type");
615 sigchain_push(SIGPIPE, SIG_IGN);
617 assert(strlen(filter_type) < LARGE_PACKET_DATA_MAX - strlen("command=\n"));
618 err = packet_write_fmt_gently(process->in, "command=%s\n", filter_type);
622 err = strlen(path) > LARGE_PACKET_DATA_MAX - strlen("pathname=\n");
624 error("path name too long for external filter");
628 err = packet_write_fmt_gently(process->in, "pathname=%s\n", path);
632 err = packet_flush_gently(process->in);
637 err = write_packetized_from_fd(fd, process->in);
639 err = write_packetized_from_buf(src, len, process->in);
643 err = subprocess_read_status(process->out, &filter_status);
647 err = strcmp(filter_status.buf, "success");
651 err = read_packetized_to_strbuf(process->out, &nbuf) < 0;
655 err = subprocess_read_status(process->out, &filter_status);
659 err = strcmp(filter_status.buf, "success");
662 sigchain_pop(SIGPIPE);
665 if (!strcmp(filter_status.buf, "error")) {
666 /* The filter signaled a problem with the file. */
667 } else if (!strcmp(filter_status.buf, "abort")) {
669 * The filter signaled a permanent problem. Don't try to filter
670 * files with the same command for the lifetime of the current
673 entry->supported_capabilities &= ~wanted_capability;
676 * Something went wrong with the protocol filter.
677 * Force shutdown and restart if another blob requires filtering.
679 error("external filter '%s' failed", cmd);
680 subprocess_stop(&subprocess_map, &entry->subprocess);
684 strbuf_swap(dst, &nbuf);
686 strbuf_release(&nbuf);
690 static struct convert_driver {
692 struct convert_driver *next;
697 } *user_convert, **user_convert_tail;
699 static int apply_filter(const char *path, const char *src, size_t len,
700 int fd, struct strbuf *dst, struct convert_driver *drv,
701 const unsigned int wanted_capability)
703 const char *cmd = NULL;
711 if ((CAP_CLEAN & wanted_capability) && !drv->process && drv->clean)
713 else if ((CAP_SMUDGE & wanted_capability) && !drv->process && drv->smudge)
717 return apply_single_file_filter(path, src, len, fd, dst, cmd);
718 else if (drv->process && *drv->process)
719 return apply_multi_file_filter(path, src, len, fd, dst, drv->process, wanted_capability);
724 static int read_convert_config(const char *var, const char *value, void *cb)
726 const char *key, *name;
728 struct convert_driver *drv;
731 * External conversion drivers are configured using
732 * "filter.<name>.variable".
734 if (parse_config_key(var, "filter", &name, &namelen, &key) < 0 || !name)
736 for (drv = user_convert; drv; drv = drv->next)
737 if (!strncmp(drv->name, name, namelen) && !drv->name[namelen])
740 drv = xcalloc(1, sizeof(struct convert_driver));
741 drv->name = xmemdupz(name, namelen);
742 *user_convert_tail = drv;
743 user_convert_tail = &(drv->next);
747 * filter.<name>.smudge and filter.<name>.clean specifies
752 * The command-line will not be interpolated in any way.
755 if (!strcmp("smudge", key))
756 return git_config_string(&drv->smudge, var, value);
758 if (!strcmp("clean", key))
759 return git_config_string(&drv->clean, var, value);
761 if (!strcmp("process", key))
762 return git_config_string(&drv->process, var, value);
764 if (!strcmp("required", key)) {
765 drv->required = git_config_bool(var, value);
772 static int count_ident(const char *cp, unsigned long size)
775 * "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$"
787 if (memcmp("Id", cp, 2))
798 * "$Id: ... "; scan up to the closing dollar sign and discard.
814 static int ident_to_git(const char *path, const char *src, size_t len,
815 struct strbuf *buf, int ident)
819 if (!ident || (src && !count_ident(src, len)))
825 /* only grow if not in place */
826 if (strbuf_avail(buf) + buf->len < len)
827 strbuf_grow(buf, len - buf->len);
830 dollar = memchr(src, '$', len);
833 memmove(dst, src, dollar + 1 - src);
834 dst += dollar + 1 - src;
835 len -= dollar + 1 - src;
838 if (len > 3 && !memcmp(src, "Id:", 3)) {
839 dollar = memchr(src + 3, '$', len - 3);
842 if (memchr(src + 3, '\n', dollar - src - 3)) {
843 /* Line break before the next dollar. */
847 memcpy(dst, "Id$", 3);
849 len -= dollar + 1 - src;
853 memmove(dst, src, len);
854 strbuf_setlen(buf, dst + len - buf->buf);
858 static int ident_to_worktree(const char *path, const char *src, size_t len,
859 struct strbuf *buf, int ident)
861 unsigned char sha1[20];
862 char *to_free = NULL, *dollar, *spc;
868 cnt = count_ident(src, len);
872 /* are we "faking" in place editing ? */
874 to_free = strbuf_detach(buf, NULL);
875 hash_sha1_file(src, len, "blob", sha1);
877 strbuf_grow(buf, len + cnt * 43);
879 /* step 1: run to the next '$' */
880 dollar = memchr(src, '$', len);
883 strbuf_add(buf, src, dollar + 1 - src);
884 len -= dollar + 1 - src;
887 /* step 2: does it looks like a bit like Id:xxx$ or Id$ ? */
888 if (len < 3 || memcmp("Id", src, 2))
891 /* step 3: skip over Id$ or Id:xxxxx$ */
895 } else if (src[2] == ':') {
897 * It's possible that an expanded Id has crept its way into the
898 * repository, we cope with that by stripping the expansion out.
899 * This is probably not a good idea, since it will cause changes
900 * on checkout, which won't go away by stash, but let's keep it
903 dollar = memchr(src + 3, '$', len - 3);
905 /* incomplete keyword, no more '$', so just quit the loop */
909 if (memchr(src + 3, '\n', dollar - src - 3)) {
910 /* Line break before the next dollar. */
914 spc = memchr(src + 4, ' ', dollar - src - 4);
915 if (spc && spc < dollar-1) {
916 /* There are spaces in unexpected places.
917 * This is probably an id from some other
918 * versioning system. Keep it for now.
923 len -= dollar + 1 - src;
926 /* it wasn't a "Id$" or "Id:xxxx$" */
930 /* step 4: substitute */
931 strbuf_addstr(buf, "Id: ");
932 strbuf_add(buf, sha1_to_hex(sha1), 40);
933 strbuf_addstr(buf, " $");
935 strbuf_add(buf, src, len);
941 static enum crlf_action git_path_check_crlf(struct attr_check_item *check)
943 const char *value = check->value;
945 if (ATTR_TRUE(value))
947 else if (ATTR_FALSE(value))
949 else if (ATTR_UNSET(value))
951 else if (!strcmp(value, "input"))
952 return CRLF_TEXT_INPUT;
953 else if (!strcmp(value, "auto"))
955 return CRLF_UNDEFINED;
958 static enum eol git_path_check_eol(struct attr_check_item *check)
960 const char *value = check->value;
962 if (ATTR_UNSET(value))
964 else if (!strcmp(value, "lf"))
966 else if (!strcmp(value, "crlf"))
971 static struct convert_driver *git_path_check_convert(struct attr_check_item *check)
973 const char *value = check->value;
974 struct convert_driver *drv;
976 if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
978 for (drv = user_convert; drv; drv = drv->next)
979 if (!strcmp(value, drv->name))
984 static int git_path_check_ident(struct attr_check_item *check)
986 const char *value = check->value;
988 return !!ATTR_TRUE(value);
992 struct convert_driver *drv;
993 enum crlf_action attr_action; /* What attr says */
994 enum crlf_action crlf_action; /* When no attr is set, use core.autocrlf */
998 static void convert_attrs(struct conv_attrs *ca, const char *path)
1000 static struct attr_check *check;
1003 check = attr_check_initl("crlf", "ident", "filter",
1004 "eol", "text", NULL);
1005 user_convert_tail = &user_convert;
1006 git_config(read_convert_config, NULL);
1009 if (!git_check_attr(path, check)) {
1010 struct attr_check_item *ccheck = check->items;
1011 ca->crlf_action = git_path_check_crlf(ccheck + 4);
1012 if (ca->crlf_action == CRLF_UNDEFINED)
1013 ca->crlf_action = git_path_check_crlf(ccheck + 0);
1014 ca->attr_action = ca->crlf_action;
1015 ca->ident = git_path_check_ident(ccheck + 1);
1016 ca->drv = git_path_check_convert(ccheck + 2);
1017 if (ca->crlf_action != CRLF_BINARY) {
1018 enum eol eol_attr = git_path_check_eol(ccheck + 3);
1019 if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_LF)
1020 ca->crlf_action = CRLF_AUTO_INPUT;
1021 else if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_CRLF)
1022 ca->crlf_action = CRLF_AUTO_CRLF;
1023 else if (eol_attr == EOL_LF)
1024 ca->crlf_action = CRLF_TEXT_INPUT;
1025 else if (eol_attr == EOL_CRLF)
1026 ca->crlf_action = CRLF_TEXT_CRLF;
1028 ca->attr_action = ca->crlf_action;
1031 ca->crlf_action = CRLF_UNDEFINED;
1034 if (ca->crlf_action == CRLF_TEXT)
1035 ca->crlf_action = text_eol_is_crlf() ? CRLF_TEXT_CRLF : CRLF_TEXT_INPUT;
1036 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_FALSE)
1037 ca->crlf_action = CRLF_BINARY;
1038 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_TRUE)
1039 ca->crlf_action = CRLF_AUTO_CRLF;
1040 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_INPUT)
1041 ca->crlf_action = CRLF_AUTO_INPUT;
1044 int would_convert_to_git_filter_fd(const char *path)
1046 struct conv_attrs ca;
1048 convert_attrs(&ca, path);
1053 * Apply a filter to an fd only if the filter is required to succeed.
1054 * We must die if the filter fails, because the original data before
1055 * filtering is not available.
1057 if (!ca.drv->required)
1060 return apply_filter(path, NULL, 0, -1, NULL, ca.drv, CAP_CLEAN);
1063 const char *get_convert_attr_ascii(const char *path)
1065 struct conv_attrs ca;
1067 convert_attrs(&ca, path);
1068 switch (ca.attr_action) {
1069 case CRLF_UNDEFINED:
1075 case CRLF_TEXT_INPUT:
1076 return "text eol=lf";
1077 case CRLF_TEXT_CRLF:
1078 return "text eol=crlf";
1081 case CRLF_AUTO_CRLF:
1082 return "text=auto eol=crlf";
1083 case CRLF_AUTO_INPUT:
1084 return "text=auto eol=lf";
1089 int convert_to_git(const struct index_state *istate,
1090 const char *path, const char *src, size_t len,
1091 struct strbuf *dst, enum safe_crlf checksafe)
1094 struct conv_attrs ca;
1096 convert_attrs(&ca, path);
1098 ret |= apply_filter(path, src, len, -1, dst, ca.drv, CAP_CLEAN);
1099 if (!ret && ca.drv && ca.drv->required)
1100 die("%s: clean filter '%s' failed", path, ca.drv->name);
1106 ret |= crlf_to_git(istate, path, src, len, dst, ca.crlf_action, checksafe);
1111 return ret | ident_to_git(path, src, len, dst, ca.ident);
1114 void convert_to_git_filter_fd(const struct index_state *istate,
1115 const char *path, int fd, struct strbuf *dst,
1116 enum safe_crlf checksafe)
1118 struct conv_attrs ca;
1119 convert_attrs(&ca, path);
1122 assert(ca.drv->clean || ca.drv->process);
1124 if (!apply_filter(path, NULL, 0, fd, dst, ca.drv, CAP_CLEAN))
1125 die("%s: clean filter '%s' failed", path, ca.drv->name);
1127 crlf_to_git(istate, path, dst->buf, dst->len, dst, ca.crlf_action, checksafe);
1128 ident_to_git(path, dst->buf, dst->len, dst, ca.ident);
1131 static int convert_to_working_tree_internal(const char *path, const char *src,
1132 size_t len, struct strbuf *dst,
1135 int ret = 0, ret_filter = 0;
1136 struct conv_attrs ca;
1138 convert_attrs(&ca, path);
1140 ret |= ident_to_worktree(path, src, len, dst, ca.ident);
1146 * CRLF conversion can be skipped if normalizing, unless there
1147 * is a smudge or process filter (even if the process filter doesn't
1148 * support smudge). The filters might expect CRLFs.
1150 if ((ca.drv && (ca.drv->smudge || ca.drv->process)) || !normalizing) {
1151 ret |= crlf_to_worktree(path, src, len, dst, ca.crlf_action);
1158 ret_filter = apply_filter(path, src, len, -1, dst, ca.drv, CAP_SMUDGE);
1159 if (!ret_filter && ca.drv && ca.drv->required)
1160 die("%s: smudge filter %s failed", path, ca.drv->name);
1162 return ret | ret_filter;
1165 int convert_to_working_tree(const char *path, const char *src, size_t len, struct strbuf *dst)
1167 return convert_to_working_tree_internal(path, src, len, dst, 0);
1170 int renormalize_buffer(const struct index_state *istate, const char *path,
1171 const char *src, size_t len, struct strbuf *dst)
1173 int ret = convert_to_working_tree_internal(path, src, len, dst, 1);
1178 return ret | convert_to_git(istate, path, src, len, dst, SAFE_CRLF_RENORMALIZE);
1181 /*****************************************************************
1183 * Streaming conversion support
1185 *****************************************************************/
1187 typedef int (*filter_fn)(struct stream_filter *,
1188 const char *input, size_t *isize_p,
1189 char *output, size_t *osize_p);
1190 typedef void (*free_fn)(struct stream_filter *);
1192 struct stream_filter_vtbl {
1197 struct stream_filter {
1198 struct stream_filter_vtbl *vtbl;
1201 static int null_filter_fn(struct stream_filter *filter,
1202 const char *input, size_t *isize_p,
1203 char *output, size_t *osize_p)
1208 return 0; /* we do not keep any states */
1210 if (*osize_p < count)
1213 memmove(output, input, count);
1220 static void null_free_fn(struct stream_filter *filter)
1222 ; /* nothing -- null instances are shared */
1225 static struct stream_filter_vtbl null_vtbl = {
1230 static struct stream_filter null_filter_singleton = {
1234 int is_null_stream_filter(struct stream_filter *filter)
1236 return filter == &null_filter_singleton;
1244 struct lf_to_crlf_filter {
1245 struct stream_filter filter;
1246 unsigned has_held:1;
1250 static int lf_to_crlf_filter_fn(struct stream_filter *filter,
1251 const char *input, size_t *isize_p,
1252 char *output, size_t *osize_p)
1254 size_t count, o = 0;
1255 struct lf_to_crlf_filter *lf_to_crlf = (struct lf_to_crlf_filter *)filter;
1258 * We may be holding onto the CR to see if it is followed by a
1259 * LF, in which case we would need to go to the main loop.
1260 * Otherwise, just emit it to the output stream.
1262 if (lf_to_crlf->has_held && (lf_to_crlf->held != '\r' || !input)) {
1263 output[o++] = lf_to_crlf->held;
1264 lf_to_crlf->has_held = 0;
1267 /* We are told to drain */
1274 if (count || lf_to_crlf->has_held) {
1278 if (lf_to_crlf->has_held) {
1280 lf_to_crlf->has_held = 0;
1283 for (i = 0; o < *osize_p && i < count; i++) {
1288 } else if (was_cr) {
1290 * Previous round saw CR and it is not followed
1291 * by a LF; emit the CR before processing the
1292 * current character.
1298 * We may have consumed the last output slot,
1299 * in which case we need to break out of this
1300 * loop; hold the current character before
1303 if (*osize_p <= o) {
1304 lf_to_crlf->has_held = 1;
1305 lf_to_crlf->held = ch;
1306 continue; /* break but increment i */
1321 if (!lf_to_crlf->has_held && was_cr) {
1322 lf_to_crlf->has_held = 1;
1323 lf_to_crlf->held = '\r';
1329 static void lf_to_crlf_free_fn(struct stream_filter *filter)
1334 static struct stream_filter_vtbl lf_to_crlf_vtbl = {
1335 lf_to_crlf_filter_fn,
1339 static struct stream_filter *lf_to_crlf_filter(void)
1341 struct lf_to_crlf_filter *lf_to_crlf = xcalloc(1, sizeof(*lf_to_crlf));
1343 lf_to_crlf->filter.vtbl = &lf_to_crlf_vtbl;
1344 return (struct stream_filter *)lf_to_crlf;
1350 #define FILTER_BUFFER 1024
1351 struct cascade_filter {
1352 struct stream_filter filter;
1353 struct stream_filter *one;
1354 struct stream_filter *two;
1355 char buf[FILTER_BUFFER];
1359 static int cascade_filter_fn(struct stream_filter *filter,
1360 const char *input, size_t *isize_p,
1361 char *output, size_t *osize_p)
1363 struct cascade_filter *cas = (struct cascade_filter *) filter;
1365 size_t sz = *osize_p;
1366 size_t to_feed, remaining;
1369 * input -- (one) --> buf -- (two) --> output
1371 while (filled < sz) {
1372 remaining = sz - filled;
1374 /* do we already have something to feed two with? */
1375 if (cas->ptr < cas->end) {
1376 to_feed = cas->end - cas->ptr;
1377 if (stream_filter(cas->two,
1378 cas->buf + cas->ptr, &to_feed,
1379 output + filled, &remaining))
1381 cas->ptr += (cas->end - cas->ptr) - to_feed;
1382 filled = sz - remaining;
1386 /* feed one from upstream and have it emit into our buffer */
1387 to_feed = input ? *isize_p : 0;
1388 if (input && !to_feed)
1390 remaining = sizeof(cas->buf);
1391 if (stream_filter(cas->one,
1393 cas->buf, &remaining))
1395 cas->end = sizeof(cas->buf) - remaining;
1398 size_t fed = *isize_p - to_feed;
1403 /* do we know that we drained one completely? */
1404 if (input || cas->end)
1407 /* tell two to drain; we have nothing more to give it */
1409 remaining = sz - filled;
1410 if (stream_filter(cas->two,
1412 output + filled, &remaining))
1414 if (remaining == (sz - filled))
1415 break; /* completely drained two */
1416 filled = sz - remaining;
1422 static void cascade_free_fn(struct stream_filter *filter)
1424 struct cascade_filter *cas = (struct cascade_filter *)filter;
1425 free_stream_filter(cas->one);
1426 free_stream_filter(cas->two);
1430 static struct stream_filter_vtbl cascade_vtbl = {
1435 static struct stream_filter *cascade_filter(struct stream_filter *one,
1436 struct stream_filter *two)
1438 struct cascade_filter *cascade;
1440 if (!one || is_null_stream_filter(one))
1442 if (!two || is_null_stream_filter(two))
1445 cascade = xmalloc(sizeof(*cascade));
1448 cascade->end = cascade->ptr = 0;
1449 cascade->filter.vtbl = &cascade_vtbl;
1450 return (struct stream_filter *)cascade;
1456 #define IDENT_DRAINING (-1)
1457 #define IDENT_SKIPPING (-2)
1458 struct ident_filter {
1459 struct stream_filter filter;
1462 char ident[45]; /* ": x40 $" */
1465 static int is_foreign_ident(const char *str)
1469 if (!skip_prefix(str, "$Id: ", &str))
1471 for (i = 0; str[i]; i++) {
1472 if (isspace(str[i]) && str[i+1] != '$')
1478 static void ident_drain(struct ident_filter *ident, char **output_p, size_t *osize_p)
1480 size_t to_drain = ident->left.len;
1482 if (*osize_p < to_drain)
1483 to_drain = *osize_p;
1485 memcpy(*output_p, ident->left.buf, to_drain);
1486 strbuf_remove(&ident->left, 0, to_drain);
1487 *output_p += to_drain;
1488 *osize_p -= to_drain;
1490 if (!ident->left.len)
1494 static int ident_filter_fn(struct stream_filter *filter,
1495 const char *input, size_t *isize_p,
1496 char *output, size_t *osize_p)
1498 struct ident_filter *ident = (struct ident_filter *)filter;
1499 static const char head[] = "$Id";
1502 /* drain upon eof */
1503 switch (ident->state) {
1505 strbuf_add(&ident->left, head, ident->state);
1506 case IDENT_SKIPPING:
1508 case IDENT_DRAINING:
1509 ident_drain(ident, &output, osize_p);
1514 while (*isize_p || (ident->state == IDENT_DRAINING)) {
1517 if (ident->state == IDENT_DRAINING) {
1518 ident_drain(ident, &output, osize_p);
1527 if (ident->state == IDENT_SKIPPING) {
1529 * Skipping until '$' or LF, but keeping them
1530 * in case it is a foreign ident.
1532 strbuf_addch(&ident->left, ch);
1533 if (ch != '\n' && ch != '$')
1535 if (ch == '$' && !is_foreign_ident(ident->left.buf)) {
1536 strbuf_setlen(&ident->left, sizeof(head) - 1);
1537 strbuf_addstr(&ident->left, ident->ident);
1539 ident->state = IDENT_DRAINING;
1543 if (ident->state < sizeof(head) &&
1544 head[ident->state] == ch) {
1550 strbuf_add(&ident->left, head, ident->state);
1551 if (ident->state == sizeof(head) - 1) {
1552 if (ch != ':' && ch != '$') {
1553 strbuf_addch(&ident->left, ch);
1559 strbuf_addch(&ident->left, ch);
1560 ident->state = IDENT_SKIPPING;
1562 strbuf_addstr(&ident->left, ident->ident);
1563 ident->state = IDENT_DRAINING;
1568 strbuf_addch(&ident->left, ch);
1569 ident->state = IDENT_DRAINING;
1574 static void ident_free_fn(struct stream_filter *filter)
1576 struct ident_filter *ident = (struct ident_filter *)filter;
1577 strbuf_release(&ident->left);
1581 static struct stream_filter_vtbl ident_vtbl = {
1586 static struct stream_filter *ident_filter(const unsigned char *sha1)
1588 struct ident_filter *ident = xmalloc(sizeof(*ident));
1590 xsnprintf(ident->ident, sizeof(ident->ident),
1591 ": %s $", sha1_to_hex(sha1));
1592 strbuf_init(&ident->left, 0);
1593 ident->filter.vtbl = &ident_vtbl;
1595 return (struct stream_filter *)ident;
1599 * Return an appropriately constructed filter for the path, or NULL if
1600 * the contents cannot be filtered without reading the whole thing
1603 * Note that you would be crazy to set CRLF, smuge/clean or ident to a
1604 * large binary blob you would want us not to slurp into the memory!
1606 struct stream_filter *get_stream_filter(const char *path, const unsigned char *sha1)
1608 struct conv_attrs ca;
1609 struct stream_filter *filter = NULL;
1611 convert_attrs(&ca, path);
1612 if (ca.drv && (ca.drv->process || ca.drv->smudge || ca.drv->clean))
1615 if (ca.crlf_action == CRLF_AUTO || ca.crlf_action == CRLF_AUTO_CRLF)
1619 filter = ident_filter(sha1);
1621 if (output_eol(ca.crlf_action) == EOL_CRLF)
1622 filter = cascade_filter(filter, lf_to_crlf_filter());
1624 filter = cascade_filter(filter, &null_filter_singleton);
1629 void free_stream_filter(struct stream_filter *filter)
1631 filter->vtbl->free(filter);
1634 int stream_filter(struct stream_filter *filter,
1635 const char *input, size_t *isize_p,
1636 char *output, size_t *osize_p)
1638 return filter->vtbl->filter(filter, input, isize_p, output, osize_p);