3 #include "run-command.h"
8 * convert.c - convert a file when checking it out and checking it in.
10 * This should use the pathname to decide on whether it wants to do some
11 * more interesting conversions (automatic gzip/unzip, general format
12 * conversions etc etc), but by default it just does automatic CRLF<->LF
13 * translation when the "text" attribute or "auto_crlf" option is set.
26 /* NUL, CR, LF and CRLF counts */
27 unsigned nul, cr, lf, crlf;
29 /* These are just approximations! */
30 unsigned printable, nonprintable;
33 static void gather_stats(const char *buf, unsigned long size, struct text_stat *stats)
37 memset(stats, 0, sizeof(*stats));
39 for (i = 0; i < size; i++) {
40 unsigned char c = buf[i];
43 if (i+1 < size && buf[i+1] == '\n')
53 stats->nonprintable++;
56 /* BS, HT, ESC and FF */
57 case '\b': case '\t': case '\033': case '\014':
64 stats->nonprintable++;
71 /* If file ends with EOF then don't count this EOF as non-printable. */
72 if (size >= 1 && buf[size-1] == '\032')
73 stats->nonprintable--;
77 * The same heuristics as diff.c::mmfile_is_binary()
79 static int is_binary(unsigned long size, struct text_stat *stats)
84 if ((stats->printable >> 7) < stats->nonprintable)
87 * Other heuristics? Average line length might be relevant,
88 * as might LF vs CR vs CRLF counts..
90 * NOTE! It might be normal to have a low ratio of CRLF to LF
91 * (somebody starts with a LF-only file and edits it with an editor
92 * that adds CRLF only to lines that are added..). But do we
93 * want to support CR-only? Probably not.
98 static enum eol output_eol(enum crlf_action crlf_action)
100 switch (crlf_action) {
113 if (auto_crlf == AUTO_CRLF_TRUE)
115 else if (auto_crlf == AUTO_CRLF_INPUT)
117 else if (core_eol == EOL_UNSET)
123 static void check_safe_crlf(const char *path, enum crlf_action crlf_action,
124 struct text_stat *stats, enum safe_crlf checksafe)
129 if (output_eol(crlf_action) == EOL_LF) {
131 * CRLFs would not be restored by checkout:
132 * check if we'd remove CRLFs
135 if (checksafe == SAFE_CRLF_WARN)
136 warning("CRLF will be replaced by LF in %s.\nThe file will have its original line endings in your working directory.", path);
137 else /* i.e. SAFE_CRLF_FAIL */
138 die("CRLF would be replaced by LF in %s.", path);
140 } else if (output_eol(crlf_action) == EOL_CRLF) {
142 * CRLFs would be added by checkout:
143 * check if we have "naked" LFs
145 if (stats->lf != stats->crlf) {
146 if (checksafe == SAFE_CRLF_WARN)
147 warning("LF will be replaced by CRLF in %s.\nThe file will have its original line endings in your working directory.", path);
148 else /* i.e. SAFE_CRLF_FAIL */
149 die("LF would be replaced by CRLF in %s", path);
154 static int has_cr_in_index(const char *path)
160 data = read_blob_data_from_cache(path, &sz);
163 has_cr = memchr(data, '\r', sz) != NULL;
168 static int crlf_to_git(const char *path, const char *src, size_t len,
170 enum crlf_action crlf_action, enum safe_crlf checksafe)
172 struct text_stat stats;
175 if (crlf_action == CRLF_BINARY ||
176 (crlf_action == CRLF_GUESS && auto_crlf == AUTO_CRLF_FALSE) ||
181 * If we are doing a dry-run and have no source buffer, there is
182 * nothing to analyze; we must assume we would convert.
187 gather_stats(src, len, &stats);
189 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_GUESS) {
191 * We're currently not going to even try to convert stuff
192 * that has bare CR characters. Does anybody do that crazy
195 if (stats.cr != stats.crlf)
199 * And add some heuristics for binary vs text, of course...
201 if (is_binary(len, &stats))
204 if (crlf_action == CRLF_GUESS) {
206 * If the file in the index has any CR in it, do not convert.
207 * This is the new safer autocrlf handling.
209 if (has_cr_in_index(path))
214 check_safe_crlf(path, crlf_action, &stats, checksafe);
216 /* Optimization: No CR? Nothing to convert, regardless. */
221 * At this point all of our source analysis is done, and we are sure we
222 * would convert. If we are in dry-run mode, we can give an answer.
227 /* only grow if not in place */
228 if (strbuf_avail(buf) + buf->len < len)
229 strbuf_grow(buf, len - buf->len);
231 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_GUESS) {
233 * If we guessed, we already know we rejected a file with
234 * lone CR, and we can strip a CR without looking at what
238 unsigned char c = *src++;
244 unsigned char c = *src++;
245 if (! (c == '\r' && (1 < len && *src == '\n')))
249 strbuf_setlen(buf, dst - buf->buf);
253 static int crlf_to_worktree(const char *path, const char *src, size_t len,
254 struct strbuf *buf, enum crlf_action crlf_action)
256 char *to_free = NULL;
257 struct text_stat stats;
259 if (!len || output_eol(crlf_action) != EOL_CRLF)
262 gather_stats(src, len, &stats);
264 /* No LF? Nothing to convert, regardless. */
268 /* Was it already in CRLF format? */
269 if (stats.lf == stats.crlf)
272 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_GUESS) {
273 if (crlf_action == CRLF_GUESS) {
274 /* If we have any CR or CRLF line endings, we do not touch it */
275 /* This is the new safer autocrlf-handling */
276 if (stats.cr > 0 || stats.crlf > 0)
280 /* If we have any bare CR characters, we're not going to touch it */
281 if (stats.cr != stats.crlf)
284 if (is_binary(len, &stats))
288 /* are we "faking" in place editing ? */
290 to_free = strbuf_detach(buf, NULL);
292 strbuf_grow(buf, len + stats.lf - stats.crlf);
294 const char *nl = memchr(src, '\n', len);
297 if (nl > src && nl[-1] == '\r') {
298 strbuf_add(buf, src, nl + 1 - src);
300 strbuf_add(buf, src, nl - src);
301 strbuf_addstr(buf, "\r\n");
306 strbuf_add(buf, src, len);
312 struct filter_params {
320 static int filter_buffer_or_fd(int in, int out, void *data)
323 * Spawn cmd and feed the buffer contents through its stdin.
325 struct child_process child_process;
326 struct filter_params *params = (struct filter_params *)data;
327 int write_err, status;
328 const char *argv[] = { NULL, NULL };
330 /* apply % substitution to cmd */
331 struct strbuf cmd = STRBUF_INIT;
332 struct strbuf path = STRBUF_INIT;
333 struct strbuf_expand_dict_entry dict[] = {
338 /* quote the path to preserve spaces, etc. */
339 sq_quote_buf(&path, params->path);
340 dict[0].value = path.buf;
342 /* expand all %f with the quoted path */
343 strbuf_expand(&cmd, params->cmd, strbuf_expand_dict_cb, &dict);
344 strbuf_release(&path);
348 memset(&child_process, 0, sizeof(child_process));
349 child_process.argv = argv;
350 child_process.use_shell = 1;
351 child_process.in = -1;
352 child_process.out = out;
354 if (start_command(&child_process))
355 return error("cannot fork to run external filter %s", params->cmd);
357 sigchain_push(SIGPIPE, SIG_IGN);
360 write_err = (write_in_full(child_process.in, params->src, params->size) < 0);
362 write_err = copy_fd(params->fd, child_process.in);
365 if (close(child_process.in))
368 error("cannot feed the input to external filter %s", params->cmd);
370 sigchain_pop(SIGPIPE);
372 status = finish_command(&child_process);
374 error("external filter %s failed %d", params->cmd, status);
376 strbuf_release(&cmd);
377 return (write_err || status);
380 static int apply_filter(const char *path, const char *src, size_t len, int fd,
381 struct strbuf *dst, const char *cmd)
384 * Create a pipeline to have the command filter the buffer's
387 * (child --> cmd) --> us
390 struct strbuf nbuf = STRBUF_INIT;
392 struct filter_params params;
400 memset(&async, 0, sizeof(async));
401 async.proc = filter_buffer_or_fd;
402 async.data = ¶ms;
411 if (start_async(&async))
412 return 0; /* error was already reported */
414 if (strbuf_read(&nbuf, async.out, len) < 0) {
415 error("read from external filter %s failed", cmd);
418 if (close(async.out)) {
419 error("read from external filter %s failed", cmd);
422 if (finish_async(&async)) {
423 error("external filter %s failed", cmd);
428 strbuf_swap(dst, &nbuf);
430 strbuf_release(&nbuf);
434 static struct convert_driver {
436 struct convert_driver *next;
440 } *user_convert, **user_convert_tail;
442 static int read_convert_config(const char *var, const char *value, void *cb)
444 const char *key, *name;
446 struct convert_driver *drv;
449 * External conversion drivers are configured using
450 * "filter.<name>.variable".
452 if (parse_config_key(var, "filter", &name, &namelen, &key) < 0 || !name)
454 for (drv = user_convert; drv; drv = drv->next)
455 if (!strncmp(drv->name, name, namelen) && !drv->name[namelen])
458 drv = xcalloc(1, sizeof(struct convert_driver));
459 drv->name = xmemdupz(name, namelen);
460 *user_convert_tail = drv;
461 user_convert_tail = &(drv->next);
465 * filter.<name>.smudge and filter.<name>.clean specifies
470 * The command-line will not be interpolated in any way.
473 if (!strcmp("smudge", key))
474 return git_config_string(&drv->smudge, var, value);
476 if (!strcmp("clean", key))
477 return git_config_string(&drv->clean, var, value);
479 if (!strcmp("required", key)) {
480 drv->required = git_config_bool(var, value);
487 static int count_ident(const char *cp, unsigned long size)
490 * "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$"
502 if (memcmp("Id", cp, 2))
513 * "$Id: ... "; scan up to the closing dollar sign and discard.
529 static int ident_to_git(const char *path, const char *src, size_t len,
530 struct strbuf *buf, int ident)
534 if (!ident || (src && !count_ident(src, len)))
540 /* only grow if not in place */
541 if (strbuf_avail(buf) + buf->len < len)
542 strbuf_grow(buf, len - buf->len);
545 dollar = memchr(src, '$', len);
548 memmove(dst, src, dollar + 1 - src);
549 dst += dollar + 1 - src;
550 len -= dollar + 1 - src;
553 if (len > 3 && !memcmp(src, "Id:", 3)) {
554 dollar = memchr(src + 3, '$', len - 3);
557 if (memchr(src + 3, '\n', dollar - src - 3)) {
558 /* Line break before the next dollar. */
562 memcpy(dst, "Id$", 3);
564 len -= dollar + 1 - src;
568 memmove(dst, src, len);
569 strbuf_setlen(buf, dst + len - buf->buf);
573 static int ident_to_worktree(const char *path, const char *src, size_t len,
574 struct strbuf *buf, int ident)
576 unsigned char sha1[20];
577 char *to_free = NULL, *dollar, *spc;
583 cnt = count_ident(src, len);
587 /* are we "faking" in place editing ? */
589 to_free = strbuf_detach(buf, NULL);
590 hash_sha1_file(src, len, "blob", sha1);
592 strbuf_grow(buf, len + cnt * 43);
594 /* step 1: run to the next '$' */
595 dollar = memchr(src, '$', len);
598 strbuf_add(buf, src, dollar + 1 - src);
599 len -= dollar + 1 - src;
602 /* step 2: does it looks like a bit like Id:xxx$ or Id$ ? */
603 if (len < 3 || memcmp("Id", src, 2))
606 /* step 3: skip over Id$ or Id:xxxxx$ */
610 } else if (src[2] == ':') {
612 * It's possible that an expanded Id has crept its way into the
613 * repository, we cope with that by stripping the expansion out.
614 * This is probably not a good idea, since it will cause changes
615 * on checkout, which won't go away by stash, but let's keep it
618 dollar = memchr(src + 3, '$', len - 3);
620 /* incomplete keyword, no more '$', so just quit the loop */
624 if (memchr(src + 3, '\n', dollar - src - 3)) {
625 /* Line break before the next dollar. */
629 spc = memchr(src + 4, ' ', dollar - src - 4);
630 if (spc && spc < dollar-1) {
631 /* There are spaces in unexpected places.
632 * This is probably an id from some other
633 * versioning system. Keep it for now.
638 len -= dollar + 1 - src;
641 /* it wasn't a "Id$" or "Id:xxxx$" */
645 /* step 4: substitute */
646 strbuf_addstr(buf, "Id: ");
647 strbuf_add(buf, sha1_to_hex(sha1), 40);
648 strbuf_addstr(buf, " $");
650 strbuf_add(buf, src, len);
656 static enum crlf_action git_path_check_crlf(const char *path, struct git_attr_check *check)
658 const char *value = check->value;
660 if (ATTR_TRUE(value))
662 else if (ATTR_FALSE(value))
664 else if (ATTR_UNSET(value))
666 else if (!strcmp(value, "input"))
668 else if (!strcmp(value, "auto"))
673 static enum eol git_path_check_eol(const char *path, struct git_attr_check *check)
675 const char *value = check->value;
677 if (ATTR_UNSET(value))
679 else if (!strcmp(value, "lf"))
681 else if (!strcmp(value, "crlf"))
686 static struct convert_driver *git_path_check_convert(const char *path,
687 struct git_attr_check *check)
689 const char *value = check->value;
690 struct convert_driver *drv;
692 if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
694 for (drv = user_convert; drv; drv = drv->next)
695 if (!strcmp(value, drv->name))
700 static int git_path_check_ident(const char *path, struct git_attr_check *check)
702 const char *value = check->value;
704 return !!ATTR_TRUE(value);
707 static enum crlf_action input_crlf_action(enum crlf_action text_attr, enum eol eol_attr)
709 if (text_attr == CRLF_BINARY)
711 if (eol_attr == EOL_LF)
713 if (eol_attr == EOL_CRLF)
719 struct convert_driver *drv;
720 enum crlf_action crlf_action;
725 static const char *conv_attr_name[] = {
726 "crlf", "ident", "filter", "eol", "text",
728 #define NUM_CONV_ATTRS ARRAY_SIZE(conv_attr_name)
730 static void convert_attrs(struct conv_attrs *ca, const char *path)
733 static struct git_attr_check ccheck[NUM_CONV_ATTRS];
735 if (!ccheck[0].attr) {
736 for (i = 0; i < NUM_CONV_ATTRS; i++)
737 ccheck[i].attr = git_attr(conv_attr_name[i]);
738 user_convert_tail = &user_convert;
739 git_config(read_convert_config, NULL);
742 if (!git_check_attr(path, NUM_CONV_ATTRS, ccheck)) {
743 ca->crlf_action = git_path_check_crlf(path, ccheck + 4);
744 if (ca->crlf_action == CRLF_GUESS)
745 ca->crlf_action = git_path_check_crlf(path, ccheck + 0);
746 ca->ident = git_path_check_ident(path, ccheck + 1);
747 ca->drv = git_path_check_convert(path, ccheck + 2);
748 ca->eol_attr = git_path_check_eol(path, ccheck + 3);
751 ca->crlf_action = CRLF_GUESS;
752 ca->eol_attr = EOL_UNSET;
757 int would_convert_to_git_filter_fd(const char *path)
759 struct conv_attrs ca;
761 convert_attrs(&ca, path);
766 * Apply a filter to an fd only if the filter is required to succeed.
767 * We must die if the filter fails, because the original data before
768 * filtering is not available.
770 if (!ca.drv->required)
773 return apply_filter(path, NULL, 0, -1, NULL, ca.drv->clean);
776 int convert_to_git(const char *path, const char *src, size_t len,
777 struct strbuf *dst, enum safe_crlf checksafe)
780 const char *filter = NULL;
782 struct conv_attrs ca;
784 convert_attrs(&ca, path);
786 filter = ca.drv->clean;
787 required = ca.drv->required;
790 ret |= apply_filter(path, src, len, -1, dst, filter);
791 if (!ret && required)
792 die("%s: clean filter '%s' failed", path, ca.drv->name);
798 ca.crlf_action = input_crlf_action(ca.crlf_action, ca.eol_attr);
799 ret |= crlf_to_git(path, src, len, dst, ca.crlf_action, checksafe);
804 return ret | ident_to_git(path, src, len, dst, ca.ident);
807 void convert_to_git_filter_fd(const char *path, int fd, struct strbuf *dst,
808 enum safe_crlf checksafe)
810 struct conv_attrs ca;
811 convert_attrs(&ca, path);
814 assert(ca.drv->clean);
816 if (!apply_filter(path, NULL, 0, fd, dst, ca.drv->clean))
817 die("%s: clean filter '%s' failed", path, ca.drv->name);
819 ca.crlf_action = input_crlf_action(ca.crlf_action, ca.eol_attr);
820 crlf_to_git(path, dst->buf, dst->len, dst, ca.crlf_action, checksafe);
821 ident_to_git(path, dst->buf, dst->len, dst, ca.ident);
824 static int convert_to_working_tree_internal(const char *path, const char *src,
825 size_t len, struct strbuf *dst,
828 int ret = 0, ret_filter = 0;
829 const char *filter = NULL;
831 struct conv_attrs ca;
833 convert_attrs(&ca, path);
835 filter = ca.drv->smudge;
836 required = ca.drv->required;
839 ret |= ident_to_worktree(path, src, len, dst, ca.ident);
845 * CRLF conversion can be skipped if normalizing, unless there
846 * is a smudge filter. The filter might expect CRLFs.
848 if (filter || !normalizing) {
849 ca.crlf_action = input_crlf_action(ca.crlf_action, ca.eol_attr);
850 ret |= crlf_to_worktree(path, src, len, dst, ca.crlf_action);
857 ret_filter = apply_filter(path, src, len, -1, dst, filter);
858 if (!ret_filter && required)
859 die("%s: smudge filter %s failed", path, ca.drv->name);
861 return ret | ret_filter;
864 int convert_to_working_tree(const char *path, const char *src, size_t len, struct strbuf *dst)
866 return convert_to_working_tree_internal(path, src, len, dst, 0);
869 int renormalize_buffer(const char *path, const char *src, size_t len, struct strbuf *dst)
871 int ret = convert_to_working_tree_internal(path, src, len, dst, 1);
876 return ret | convert_to_git(path, src, len, dst, SAFE_CRLF_FALSE);
879 /*****************************************************************
881 * Streaming conversion support
883 *****************************************************************/
885 typedef int (*filter_fn)(struct stream_filter *,
886 const char *input, size_t *isize_p,
887 char *output, size_t *osize_p);
888 typedef void (*free_fn)(struct stream_filter *);
890 struct stream_filter_vtbl {
895 struct stream_filter {
896 struct stream_filter_vtbl *vtbl;
899 static int null_filter_fn(struct stream_filter *filter,
900 const char *input, size_t *isize_p,
901 char *output, size_t *osize_p)
906 return 0; /* we do not keep any states */
908 if (*osize_p < count)
911 memmove(output, input, count);
918 static void null_free_fn(struct stream_filter *filter)
920 ; /* nothing -- null instances are shared */
923 static struct stream_filter_vtbl null_vtbl = {
928 static struct stream_filter null_filter_singleton = {
932 int is_null_stream_filter(struct stream_filter *filter)
934 return filter == &null_filter_singleton;
942 struct lf_to_crlf_filter {
943 struct stream_filter filter;
948 static int lf_to_crlf_filter_fn(struct stream_filter *filter,
949 const char *input, size_t *isize_p,
950 char *output, size_t *osize_p)
953 struct lf_to_crlf_filter *lf_to_crlf = (struct lf_to_crlf_filter *)filter;
956 * We may be holding onto the CR to see if it is followed by a
957 * LF, in which case we would need to go to the main loop.
958 * Otherwise, just emit it to the output stream.
960 if (lf_to_crlf->has_held && (lf_to_crlf->held != '\r' || !input)) {
961 output[o++] = lf_to_crlf->held;
962 lf_to_crlf->has_held = 0;
965 /* We are told to drain */
972 if (count || lf_to_crlf->has_held) {
976 if (lf_to_crlf->has_held) {
978 lf_to_crlf->has_held = 0;
981 for (i = 0; o < *osize_p && i < count; i++) {
988 * Previous round saw CR and it is not followed
989 * by a LF; emit the CR before processing the
996 * We may have consumed the last output slot,
997 * in which case we need to break out of this
998 * loop; hold the current character before
1001 if (*osize_p <= o) {
1002 lf_to_crlf->has_held = 1;
1003 lf_to_crlf->held = ch;
1004 continue; /* break but increment i */
1019 if (!lf_to_crlf->has_held && was_cr) {
1020 lf_to_crlf->has_held = 1;
1021 lf_to_crlf->held = '\r';
1027 static void lf_to_crlf_free_fn(struct stream_filter *filter)
1032 static struct stream_filter_vtbl lf_to_crlf_vtbl = {
1033 lf_to_crlf_filter_fn,
1037 static struct stream_filter *lf_to_crlf_filter(void)
1039 struct lf_to_crlf_filter *lf_to_crlf = xcalloc(1, sizeof(*lf_to_crlf));
1041 lf_to_crlf->filter.vtbl = &lf_to_crlf_vtbl;
1042 return (struct stream_filter *)lf_to_crlf;
1048 #define FILTER_BUFFER 1024
1049 struct cascade_filter {
1050 struct stream_filter filter;
1051 struct stream_filter *one;
1052 struct stream_filter *two;
1053 char buf[FILTER_BUFFER];
1057 static int cascade_filter_fn(struct stream_filter *filter,
1058 const char *input, size_t *isize_p,
1059 char *output, size_t *osize_p)
1061 struct cascade_filter *cas = (struct cascade_filter *) filter;
1063 size_t sz = *osize_p;
1064 size_t to_feed, remaining;
1067 * input -- (one) --> buf -- (two) --> output
1069 while (filled < sz) {
1070 remaining = sz - filled;
1072 /* do we already have something to feed two with? */
1073 if (cas->ptr < cas->end) {
1074 to_feed = cas->end - cas->ptr;
1075 if (stream_filter(cas->two,
1076 cas->buf + cas->ptr, &to_feed,
1077 output + filled, &remaining))
1079 cas->ptr += (cas->end - cas->ptr) - to_feed;
1080 filled = sz - remaining;
1084 /* feed one from upstream and have it emit into our buffer */
1085 to_feed = input ? *isize_p : 0;
1086 if (input && !to_feed)
1088 remaining = sizeof(cas->buf);
1089 if (stream_filter(cas->one,
1091 cas->buf, &remaining))
1093 cas->end = sizeof(cas->buf) - remaining;
1096 size_t fed = *isize_p - to_feed;
1101 /* do we know that we drained one completely? */
1102 if (input || cas->end)
1105 /* tell two to drain; we have nothing more to give it */
1107 remaining = sz - filled;
1108 if (stream_filter(cas->two,
1110 output + filled, &remaining))
1112 if (remaining == (sz - filled))
1113 break; /* completely drained two */
1114 filled = sz - remaining;
1120 static void cascade_free_fn(struct stream_filter *filter)
1122 struct cascade_filter *cas = (struct cascade_filter *)filter;
1123 free_stream_filter(cas->one);
1124 free_stream_filter(cas->two);
1128 static struct stream_filter_vtbl cascade_vtbl = {
1133 static struct stream_filter *cascade_filter(struct stream_filter *one,
1134 struct stream_filter *two)
1136 struct cascade_filter *cascade;
1138 if (!one || is_null_stream_filter(one))
1140 if (!two || is_null_stream_filter(two))
1143 cascade = xmalloc(sizeof(*cascade));
1146 cascade->end = cascade->ptr = 0;
1147 cascade->filter.vtbl = &cascade_vtbl;
1148 return (struct stream_filter *)cascade;
1154 #define IDENT_DRAINING (-1)
1155 #define IDENT_SKIPPING (-2)
1156 struct ident_filter {
1157 struct stream_filter filter;
1160 char ident[45]; /* ": x40 $" */
1163 static int is_foreign_ident(const char *str)
1167 if (!skip_prefix(str, "$Id: ", &str))
1169 for (i = 0; str[i]; i++) {
1170 if (isspace(str[i]) && str[i+1] != '$')
1176 static void ident_drain(struct ident_filter *ident, char **output_p, size_t *osize_p)
1178 size_t to_drain = ident->left.len;
1180 if (*osize_p < to_drain)
1181 to_drain = *osize_p;
1183 memcpy(*output_p, ident->left.buf, to_drain);
1184 strbuf_remove(&ident->left, 0, to_drain);
1185 *output_p += to_drain;
1186 *osize_p -= to_drain;
1188 if (!ident->left.len)
1192 static int ident_filter_fn(struct stream_filter *filter,
1193 const char *input, size_t *isize_p,
1194 char *output, size_t *osize_p)
1196 struct ident_filter *ident = (struct ident_filter *)filter;
1197 static const char head[] = "$Id";
1200 /* drain upon eof */
1201 switch (ident->state) {
1203 strbuf_add(&ident->left, head, ident->state);
1204 case IDENT_SKIPPING:
1206 case IDENT_DRAINING:
1207 ident_drain(ident, &output, osize_p);
1212 while (*isize_p || (ident->state == IDENT_DRAINING)) {
1215 if (ident->state == IDENT_DRAINING) {
1216 ident_drain(ident, &output, osize_p);
1225 if (ident->state == IDENT_SKIPPING) {
1227 * Skipping until '$' or LF, but keeping them
1228 * in case it is a foreign ident.
1230 strbuf_addch(&ident->left, ch);
1231 if (ch != '\n' && ch != '$')
1233 if (ch == '$' && !is_foreign_ident(ident->left.buf)) {
1234 strbuf_setlen(&ident->left, sizeof(head) - 1);
1235 strbuf_addstr(&ident->left, ident->ident);
1237 ident->state = IDENT_DRAINING;
1241 if (ident->state < sizeof(head) &&
1242 head[ident->state] == ch) {
1248 strbuf_add(&ident->left, head, ident->state);
1249 if (ident->state == sizeof(head) - 1) {
1250 if (ch != ':' && ch != '$') {
1251 strbuf_addch(&ident->left, ch);
1257 strbuf_addch(&ident->left, ch);
1258 ident->state = IDENT_SKIPPING;
1260 strbuf_addstr(&ident->left, ident->ident);
1261 ident->state = IDENT_DRAINING;
1266 strbuf_addch(&ident->left, ch);
1267 ident->state = IDENT_DRAINING;
1272 static void ident_free_fn(struct stream_filter *filter)
1274 struct ident_filter *ident = (struct ident_filter *)filter;
1275 strbuf_release(&ident->left);
1279 static struct stream_filter_vtbl ident_vtbl = {
1284 static struct stream_filter *ident_filter(const unsigned char *sha1)
1286 struct ident_filter *ident = xmalloc(sizeof(*ident));
1288 sprintf(ident->ident, ": %s $", sha1_to_hex(sha1));
1289 strbuf_init(&ident->left, 0);
1290 ident->filter.vtbl = &ident_vtbl;
1292 return (struct stream_filter *)ident;
1296 * Return an appropriately constructed filter for the path, or NULL if
1297 * the contents cannot be filtered without reading the whole thing
1300 * Note that you would be crazy to set CRLF, smuge/clean or ident to a
1301 * large binary blob you would want us not to slurp into the memory!
1303 struct stream_filter *get_stream_filter(const char *path, const unsigned char *sha1)
1305 struct conv_attrs ca;
1306 enum crlf_action crlf_action;
1307 struct stream_filter *filter = NULL;
1309 convert_attrs(&ca, path);
1311 if (ca.drv && (ca.drv->smudge || ca.drv->clean))
1315 filter = ident_filter(sha1);
1317 crlf_action = input_crlf_action(ca.crlf_action, ca.eol_attr);
1319 if ((crlf_action == CRLF_BINARY) || (crlf_action == CRLF_INPUT) ||
1320 (crlf_action == CRLF_GUESS && auto_crlf == AUTO_CRLF_FALSE))
1321 filter = cascade_filter(filter, &null_filter_singleton);
1323 else if (output_eol(crlf_action) == EOL_CRLF &&
1324 !(crlf_action == CRLF_AUTO || crlf_action == CRLF_GUESS))
1325 filter = cascade_filter(filter, lf_to_crlf_filter());
1330 void free_stream_filter(struct stream_filter *filter)
1332 filter->vtbl->free(filter);
1335 int stream_filter(struct stream_filter *filter,
1336 const char *input, size_t *isize_p,
1337 char *output, size_t *osize_p)
1339 return filter->vtbl->filter(filter, input, isize_p, output, osize_p);