3 #include "run-command.h"
9 * convert.c - convert a file when checking it out and checking it in.
11 * This should use the pathname to decide on whether it wants to do some
12 * more interesting conversions (automatic gzip/unzip, general format
13 * conversions etc etc), but by default it just does automatic CRLF<->LF
14 * translation when the "text" attribute or "auto_crlf" option is set.
17 /* Stat bits: When BIN is set, the txt bits are unset */
18 #define CONVERT_STAT_BITS_TXT_LF 0x1
19 #define CONVERT_STAT_BITS_TXT_CRLF 0x2
20 #define CONVERT_STAT_BITS_BIN 0x4
34 /* NUL, CR, LF and CRLF counts */
35 unsigned nul
, lonecr
, lonelf
, crlf
;
37 /* These are just approximations! */
38 unsigned printable
, nonprintable
;
41 static void gather_stats(const char *buf
, unsigned long size
, struct text_stat
*stats
)
45 memset(stats
, 0, sizeof(*stats
));
47 for (i
= 0; i
< size
; i
++) {
48 unsigned char c
= buf
[i
];
50 if (i
+1 < size
&& buf
[i
+1] == '\n') {
63 stats
->nonprintable
++;
66 /* BS, HT, ESC and FF */
67 case '\b': case '\t': case '\033': case '\014':
74 stats
->nonprintable
++;
81 /* If file ends with EOF then don't count this EOF as non-printable. */
82 if (size
>= 1 && buf
[size
-1] == '\032')
83 stats
->nonprintable
--;
87 * The same heuristics as diff.c::mmfile_is_binary()
88 * We treat files with bare CR as binary
90 static int convert_is_binary(unsigned long size
, const struct text_stat
*stats
)
96 if ((stats
->printable
>> 7) < stats
->nonprintable
)
101 static unsigned int gather_convert_stats(const char *data
, unsigned long size
)
103 struct text_stat stats
;
107 gather_stats(data
, size
, &stats
);
108 if (convert_is_binary(size
, &stats
))
109 ret
|= CONVERT_STAT_BITS_BIN
;
111 ret
|= CONVERT_STAT_BITS_TXT_CRLF
;
113 ret
|= CONVERT_STAT_BITS_TXT_LF
;
118 static const char *gather_convert_stats_ascii(const char *data
, unsigned long size
)
120 unsigned int convert_stats
= gather_convert_stats(data
, size
);
122 if (convert_stats
& CONVERT_STAT_BITS_BIN
)
124 switch (convert_stats
) {
125 case CONVERT_STAT_BITS_TXT_LF
:
127 case CONVERT_STAT_BITS_TXT_CRLF
:
129 case CONVERT_STAT_BITS_TXT_LF
| CONVERT_STAT_BITS_TXT_CRLF
:
136 const char *get_cached_convert_stats_ascii(const char *path
)
140 void *data
= read_blob_data_from_cache(path
, &sz
);
141 ret
= gather_convert_stats_ascii(data
, sz
);
146 const char *get_wt_convert_stats_ascii(const char *path
)
148 const char *ret
= "";
149 struct strbuf sb
= STRBUF_INIT
;
150 if (strbuf_read_file(&sb
, path
, 0) >= 0)
151 ret
= gather_convert_stats_ascii(sb
.buf
, sb
.len
);
156 static int text_eol_is_crlf(void)
158 if (auto_crlf
== AUTO_CRLF_TRUE
)
160 else if (auto_crlf
== AUTO_CRLF_INPUT
)
162 if (core_eol
== EOL_CRLF
)
164 if (core_eol
== EOL_UNSET
&& EOL_NATIVE
== EOL_CRLF
)
169 static enum eol
output_eol(enum crlf_action crlf_action
)
171 switch (crlf_action
) {
176 case CRLF_TEXT_INPUT
:
181 case CRLF_AUTO_INPUT
:
186 return text_eol_is_crlf() ? EOL_CRLF
: EOL_LF
;
188 warning("Illegal crlf_action %d\n", (int)crlf_action
);
192 static void check_safe_crlf(const char *path
, enum crlf_action crlf_action
,
193 struct text_stat
*old_stats
, struct text_stat
*new_stats
,
194 enum safe_crlf checksafe
)
196 if (old_stats
->crlf
&& !new_stats
->crlf
) {
198 * CRLFs would not be restored by checkout
200 if (checksafe
== SAFE_CRLF_WARN
)
201 warning(_("CRLF will be replaced by LF in %s.\n"
202 "The file will have its original line"
203 " endings in your working directory."), path
);
204 else /* i.e. SAFE_CRLF_FAIL */
205 die(_("CRLF would be replaced by LF in %s."), path
);
206 } else if (old_stats
->lonelf
&& !new_stats
->lonelf
) {
208 * CRLFs would be added by checkout
210 if (checksafe
== SAFE_CRLF_WARN
)
211 warning(_("LF will be replaced by CRLF in %s.\n"
212 "The file will have its original line"
213 " endings in your working directory."), path
);
214 else /* i.e. SAFE_CRLF_FAIL */
215 die(_("LF would be replaced by CRLF in %s"), path
);
219 static int has_cr_in_index(const char *path
)
225 data
= read_blob_data_from_cache(path
, &sz
);
228 has_cr
= memchr(data
, '\r', sz
) != NULL
;
233 static int will_convert_lf_to_crlf(size_t len
, struct text_stat
*stats
,
234 enum crlf_action crlf_action
)
236 if (output_eol(crlf_action
) != EOL_CRLF
)
238 /* No "naked" LF? Nothing to convert, regardless. */
242 if (crlf_action
== CRLF_AUTO
|| crlf_action
== CRLF_AUTO_INPUT
|| crlf_action
== CRLF_AUTO_CRLF
) {
243 /* If we have any CR or CRLF line endings, we do not touch it */
244 /* This is the new safer autocrlf-handling */
245 if (stats
->lonecr
|| stats
->crlf
)
248 if (convert_is_binary(len
, stats
))
255 static int crlf_to_git(const char *path
, const char *src
, size_t len
,
257 enum crlf_action crlf_action
, enum safe_crlf checksafe
)
259 struct text_stat stats
;
261 int convert_crlf_into_lf
;
263 if (crlf_action
== CRLF_BINARY
||
268 * If we are doing a dry-run and have no source buffer, there is
269 * nothing to analyze; we must assume we would convert.
274 gather_stats(src
, len
, &stats
);
275 /* Optimization: No CRLF? Nothing to convert, regardless. */
276 convert_crlf_into_lf
= !!stats
.crlf
;
278 if (crlf_action
== CRLF_AUTO
|| crlf_action
== CRLF_AUTO_INPUT
|| crlf_action
== CRLF_AUTO_CRLF
) {
279 if (convert_is_binary(len
, &stats
))
282 * If the file in the index has any CR in it, do not
283 * convert. This is the new safer autocrlf handling,
284 * unless we want to renormalize in a merge or
287 if ((checksafe
!= SAFE_CRLF_RENORMALIZE
) && has_cr_in_index(path
))
288 convert_crlf_into_lf
= 0;
290 if ((checksafe
== SAFE_CRLF_WARN
||
291 (checksafe
== SAFE_CRLF_FAIL
)) && len
) {
292 struct text_stat new_stats
;
293 memcpy(&new_stats
, &stats
, sizeof(new_stats
));
294 /* simulate "git add" */
295 if (convert_crlf_into_lf
) {
296 new_stats
.lonelf
+= new_stats
.crlf
;
299 /* simulate "git checkout" */
300 if (will_convert_lf_to_crlf(len
, &new_stats
, crlf_action
)) {
301 new_stats
.crlf
+= new_stats
.lonelf
;
302 new_stats
.lonelf
= 0;
304 check_safe_crlf(path
, crlf_action
, &stats
, &new_stats
, checksafe
);
306 if (!convert_crlf_into_lf
)
310 * At this point all of our source analysis is done, and we are sure we
311 * would convert. If we are in dry-run mode, we can give an answer.
316 /* only grow if not in place */
317 if (strbuf_avail(buf
) + buf
->len
< len
)
318 strbuf_grow(buf
, len
- buf
->len
);
320 if (crlf_action
== CRLF_AUTO
|| crlf_action
== CRLF_AUTO_INPUT
|| crlf_action
== CRLF_AUTO_CRLF
) {
322 * If we guessed, we already know we rejected a file with
323 * lone CR, and we can strip a CR without looking at what
327 unsigned char c
= *src
++;
333 unsigned char c
= *src
++;
334 if (! (c
== '\r' && (1 < len
&& *src
== '\n')))
338 strbuf_setlen(buf
, dst
- buf
->buf
);
342 static int crlf_to_worktree(const char *path
, const char *src
, size_t len
,
343 struct strbuf
*buf
, enum crlf_action crlf_action
)
345 char *to_free
= NULL
;
346 struct text_stat stats
;
348 if (!len
|| output_eol(crlf_action
) != EOL_CRLF
)
351 gather_stats(src
, len
, &stats
);
352 if (!will_convert_lf_to_crlf(len
, &stats
, crlf_action
))
355 /* are we "faking" in place editing ? */
357 to_free
= strbuf_detach(buf
, NULL
);
359 strbuf_grow(buf
, len
+ stats
.lonelf
);
361 const char *nl
= memchr(src
, '\n', len
);
364 if (nl
> src
&& nl
[-1] == '\r') {
365 strbuf_add(buf
, src
, nl
+ 1 - src
);
367 strbuf_add(buf
, src
, nl
- src
);
368 strbuf_addstr(buf
, "\r\n");
373 strbuf_add(buf
, src
, len
);
379 struct filter_params
{
387 static int filter_buffer_or_fd(int in
, int out
, void *data
)
390 * Spawn cmd and feed the buffer contents through its stdin.
392 struct child_process child_process
= CHILD_PROCESS_INIT
;
393 struct filter_params
*params
= (struct filter_params
*)data
;
394 int write_err
, status
;
395 const char *argv
[] = { NULL
, NULL
};
397 /* apply % substitution to cmd */
398 struct strbuf cmd
= STRBUF_INIT
;
399 struct strbuf path
= STRBUF_INIT
;
400 struct strbuf_expand_dict_entry dict
[] = {
405 /* quote the path to preserve spaces, etc. */
406 sq_quote_buf(&path
, params
->path
);
407 dict
[0].value
= path
.buf
;
409 /* expand all %f with the quoted path */
410 strbuf_expand(&cmd
, params
->cmd
, strbuf_expand_dict_cb
, &dict
);
411 strbuf_release(&path
);
415 child_process
.argv
= argv
;
416 child_process
.use_shell
= 1;
417 child_process
.in
= -1;
418 child_process
.out
= out
;
420 if (start_command(&child_process
))
421 return error("cannot fork to run external filter '%s'", params
->cmd
);
423 sigchain_push(SIGPIPE
, SIG_IGN
);
426 write_err
= (write_in_full(child_process
.in
,
427 params
->src
, params
->size
) < 0);
431 write_err
= copy_fd(params
->fd
, child_process
.in
);
432 if (write_err
== COPY_WRITE_ERROR
&& errno
== EPIPE
)
436 if (close(child_process
.in
))
439 error("cannot feed the input to external filter '%s'", params
->cmd
);
441 sigchain_pop(SIGPIPE
);
443 status
= finish_command(&child_process
);
445 error("external filter '%s' failed %d", params
->cmd
, status
);
447 strbuf_release(&cmd
);
448 return (write_err
|| status
);
451 static int apply_single_file_filter(const char *path
, const char *src
, size_t len
, int fd
,
452 struct strbuf
*dst
, const char *cmd
)
455 * Create a pipeline to have the command filter the buffer's
458 * (child --> cmd) --> us
461 struct strbuf nbuf
= STRBUF_INIT
;
463 struct filter_params params
;
465 memset(&async
, 0, sizeof(async
));
466 async
.proc
= filter_buffer_or_fd
;
467 async
.data
= ¶ms
;
476 if (start_async(&async
))
477 return 0; /* error was already reported */
479 if (strbuf_read(&nbuf
, async
.out
, len
) < 0) {
480 err
= error("read from external filter '%s' failed", cmd
);
482 if (close(async
.out
)) {
483 err
= error("read from external filter '%s' failed", cmd
);
485 if (finish_async(&async
)) {
486 err
= error("external filter '%s' failed", cmd
);
490 strbuf_swap(dst
, &nbuf
);
492 strbuf_release(&nbuf
);
496 #define CAP_CLEAN (1u<<0)
497 #define CAP_SMUDGE (1u<<1)
500 struct hashmap_entry ent
; /* must be the first member! */
501 unsigned int supported_capabilities
;
503 struct child_process process
;
506 static int cmd_process_map_initialized
;
507 static struct hashmap cmd_process_map
;
509 static int cmd2process_cmp(const struct cmd2process
*e1
,
510 const struct cmd2process
*e2
,
513 return strcmp(e1
->cmd
, e2
->cmd
);
516 static struct cmd2process
*find_multi_file_filter_entry(struct hashmap
*hashmap
, const char *cmd
)
518 struct cmd2process key
;
519 hashmap_entry_init(&key
, strhash(cmd
));
521 return hashmap_get(hashmap
, &key
, NULL
);
524 static void read_multi_file_filter_status(int fd
, struct strbuf
*status
)
526 struct strbuf
**pair
;
529 line
= packet_read_line(fd
, NULL
);
532 pair
= strbuf_split_str(line
, '=', 2);
533 if (pair
[0] && pair
[0]->len
&& pair
[1]) {
534 /* the last "status=<foo>" line wins */
535 if (!strcmp(pair
[0]->buf
, "status=")) {
536 strbuf_reset(status
);
537 strbuf_addbuf(status
, pair
[1]);
540 strbuf_list_free(pair
);
544 static void kill_multi_file_filter(struct hashmap
*hashmap
, struct cmd2process
*entry
)
549 entry
->process
.clean_on_exit
= 0;
550 kill(entry
->process
.pid
, SIGTERM
);
551 finish_command(&entry
->process
);
553 hashmap_remove(hashmap
, entry
, NULL
);
557 static void stop_multi_file_filter(struct child_process
*process
)
559 sigchain_push(SIGPIPE
, SIG_IGN
);
560 /* Closing the pipe signals the filter to initiate a shutdown. */
563 sigchain_pop(SIGPIPE
);
564 /* Finish command will wait until the shutdown is complete. */
565 finish_command(process
);
568 static struct cmd2process
*start_multi_file_filter(struct hashmap
*hashmap
, const char *cmd
)
571 struct cmd2process
*entry
;
572 struct child_process
*process
;
573 const char *argv
[] = { cmd
, NULL
};
574 struct string_list cap_list
= STRING_LIST_INIT_NODUP
;
576 const char *cap_name
;
578 entry
= xmalloc(sizeof(*entry
));
580 entry
->supported_capabilities
= 0;
581 process
= &entry
->process
;
583 child_process_init(process
);
584 process
->argv
= argv
;
585 process
->use_shell
= 1;
588 process
->clean_on_exit
= 1;
589 process
->clean_on_exit_handler
= stop_multi_file_filter
;
591 if (start_command(process
)) {
592 error("cannot fork to run external filter '%s'", cmd
);
596 hashmap_entry_init(entry
, strhash(cmd
));
598 sigchain_push(SIGPIPE
, SIG_IGN
);
600 err
= packet_writel(process
->in
, "git-filter-client", "version=2", NULL
);
604 err
= strcmp(packet_read_line(process
->out
, NULL
), "git-filter-server");
606 error("external filter '%s' does not support filter protocol version 2", cmd
);
609 err
= strcmp(packet_read_line(process
->out
, NULL
), "version=2");
612 err
= packet_read_line(process
->out
, NULL
) != NULL
;
616 err
= packet_writel(process
->in
, "capability=clean", "capability=smudge", NULL
);
619 cap_buf
= packet_read_line(process
->out
, NULL
);
622 string_list_split_in_place(&cap_list
, cap_buf
, '=', 1);
624 if (cap_list
.nr
!= 2 || strcmp(cap_list
.items
[0].string
, "capability"))
627 cap_name
= cap_list
.items
[1].string
;
628 if (!strcmp(cap_name
, "clean")) {
629 entry
->supported_capabilities
|= CAP_CLEAN
;
630 } else if (!strcmp(cap_name
, "smudge")) {
631 entry
->supported_capabilities
|= CAP_SMUDGE
;
634 "external filter '%s' requested unsupported filter capability '%s'",
639 string_list_clear(&cap_list
, 0);
643 sigchain_pop(SIGPIPE
);
646 error("initialization for external filter '%s' failed", cmd
);
647 kill_multi_file_filter(hashmap
, entry
);
651 hashmap_add(hashmap
, entry
);
655 static int apply_multi_file_filter(const char *path
, const char *src
, size_t len
,
656 int fd
, struct strbuf
*dst
, const char *cmd
,
657 const unsigned int wanted_capability
)
660 struct cmd2process
*entry
;
661 struct child_process
*process
;
662 struct strbuf nbuf
= STRBUF_INIT
;
663 struct strbuf filter_status
= STRBUF_INIT
;
664 const char *filter_type
;
666 if (!cmd_process_map_initialized
) {
667 cmd_process_map_initialized
= 1;
668 hashmap_init(&cmd_process_map
, (hashmap_cmp_fn
) cmd2process_cmp
, 0);
671 entry
= find_multi_file_filter_entry(&cmd_process_map
, cmd
);
677 entry
= start_multi_file_filter(&cmd_process_map
, cmd
);
681 process
= &entry
->process
;
683 if (!(wanted_capability
& entry
->supported_capabilities
))
686 if (CAP_CLEAN
& wanted_capability
)
687 filter_type
= "clean";
688 else if (CAP_SMUDGE
& wanted_capability
)
689 filter_type
= "smudge";
691 die("unexpected filter type");
693 sigchain_push(SIGPIPE
, SIG_IGN
);
695 assert(strlen(filter_type
) < LARGE_PACKET_DATA_MAX
- strlen("command=\n"));
696 err
= packet_write_fmt_gently(process
->in
, "command=%s\n", filter_type
);
700 err
= strlen(path
) > LARGE_PACKET_DATA_MAX
- strlen("pathname=\n");
702 error("path name too long for external filter");
706 err
= packet_write_fmt_gently(process
->in
, "pathname=%s\n", path
);
710 err
= packet_flush_gently(process
->in
);
715 err
= write_packetized_from_fd(fd
, process
->in
);
717 err
= write_packetized_from_buf(src
, len
, process
->in
);
721 read_multi_file_filter_status(process
->out
, &filter_status
);
722 err
= strcmp(filter_status
.buf
, "success");
726 err
= read_packetized_to_strbuf(process
->out
, &nbuf
) < 0;
730 read_multi_file_filter_status(process
->out
, &filter_status
);
731 err
= strcmp(filter_status
.buf
, "success");
734 sigchain_pop(SIGPIPE
);
737 if (!strcmp(filter_status
.buf
, "error")) {
738 /* The filter signaled a problem with the file. */
739 } else if (!strcmp(filter_status
.buf
, "abort")) {
741 * The filter signaled a permanent problem. Don't try to filter
742 * files with the same command for the lifetime of the current
745 entry
->supported_capabilities
&= ~wanted_capability
;
748 * Something went wrong with the protocol filter.
749 * Force shutdown and restart if another blob requires filtering.
751 error("external filter '%s' failed", cmd
);
752 kill_multi_file_filter(&cmd_process_map
, entry
);
755 strbuf_swap(dst
, &nbuf
);
757 strbuf_release(&nbuf
);
761 static struct convert_driver
{
763 struct convert_driver
*next
;
768 } *user_convert
, **user_convert_tail
;
770 static int apply_filter(const char *path
, const char *src
, size_t len
,
771 int fd
, struct strbuf
*dst
, struct convert_driver
*drv
,
772 const unsigned int wanted_capability
)
774 const char *cmd
= NULL
;
782 if ((CAP_CLEAN
& wanted_capability
) && !drv
->process
&& drv
->clean
)
784 else if ((CAP_SMUDGE
& wanted_capability
) && !drv
->process
&& drv
->smudge
)
788 return apply_single_file_filter(path
, src
, len
, fd
, dst
, cmd
);
789 else if (drv
->process
&& *drv
->process
)
790 return apply_multi_file_filter(path
, src
, len
, fd
, dst
, drv
->process
, wanted_capability
);
795 static int read_convert_config(const char *var
, const char *value
, void *cb
)
797 const char *key
, *name
;
799 struct convert_driver
*drv
;
802 * External conversion drivers are configured using
803 * "filter.<name>.variable".
805 if (parse_config_key(var
, "filter", &name
, &namelen
, &key
) < 0 || !name
)
807 for (drv
= user_convert
; drv
; drv
= drv
->next
)
808 if (!strncmp(drv
->name
, name
, namelen
) && !drv
->name
[namelen
])
811 drv
= xcalloc(1, sizeof(struct convert_driver
));
812 drv
->name
= xmemdupz(name
, namelen
);
813 *user_convert_tail
= drv
;
814 user_convert_tail
= &(drv
->next
);
818 * filter.<name>.smudge and filter.<name>.clean specifies
823 * The command-line will not be interpolated in any way.
826 if (!strcmp("smudge", key
))
827 return git_config_string(&drv
->smudge
, var
, value
);
829 if (!strcmp("clean", key
))
830 return git_config_string(&drv
->clean
, var
, value
);
832 if (!strcmp("process", key
))
833 return git_config_string(&drv
->process
, var
, value
);
835 if (!strcmp("required", key
)) {
836 drv
->required
= git_config_bool(var
, value
);
843 static int count_ident(const char *cp
, unsigned long size
)
846 * "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$"
858 if (memcmp("Id", cp
, 2))
869 * "$Id: ... "; scan up to the closing dollar sign and discard.
885 static int ident_to_git(const char *path
, const char *src
, size_t len
,
886 struct strbuf
*buf
, int ident
)
890 if (!ident
|| (src
&& !count_ident(src
, len
)))
896 /* only grow if not in place */
897 if (strbuf_avail(buf
) + buf
->len
< len
)
898 strbuf_grow(buf
, len
- buf
->len
);
901 dollar
= memchr(src
, '$', len
);
904 memmove(dst
, src
, dollar
+ 1 - src
);
905 dst
+= dollar
+ 1 - src
;
906 len
-= dollar
+ 1 - src
;
909 if (len
> 3 && !memcmp(src
, "Id:", 3)) {
910 dollar
= memchr(src
+ 3, '$', len
- 3);
913 if (memchr(src
+ 3, '\n', dollar
- src
- 3)) {
914 /* Line break before the next dollar. */
918 memcpy(dst
, "Id$", 3);
920 len
-= dollar
+ 1 - src
;
924 memmove(dst
, src
, len
);
925 strbuf_setlen(buf
, dst
+ len
- buf
->buf
);
929 static int ident_to_worktree(const char *path
, const char *src
, size_t len
,
930 struct strbuf
*buf
, int ident
)
932 unsigned char sha1
[20];
933 char *to_free
= NULL
, *dollar
, *spc
;
939 cnt
= count_ident(src
, len
);
943 /* are we "faking" in place editing ? */
945 to_free
= strbuf_detach(buf
, NULL
);
946 hash_sha1_file(src
, len
, "blob", sha1
);
948 strbuf_grow(buf
, len
+ cnt
* 43);
950 /* step 1: run to the next '$' */
951 dollar
= memchr(src
, '$', len
);
954 strbuf_add(buf
, src
, dollar
+ 1 - src
);
955 len
-= dollar
+ 1 - src
;
958 /* step 2: does it looks like a bit like Id:xxx$ or Id$ ? */
959 if (len
< 3 || memcmp("Id", src
, 2))
962 /* step 3: skip over Id$ or Id:xxxxx$ */
966 } else if (src
[2] == ':') {
968 * It's possible that an expanded Id has crept its way into the
969 * repository, we cope with that by stripping the expansion out.
970 * This is probably not a good idea, since it will cause changes
971 * on checkout, which won't go away by stash, but let's keep it
974 dollar
= memchr(src
+ 3, '$', len
- 3);
976 /* incomplete keyword, no more '$', so just quit the loop */
980 if (memchr(src
+ 3, '\n', dollar
- src
- 3)) {
981 /* Line break before the next dollar. */
985 spc
= memchr(src
+ 4, ' ', dollar
- src
- 4);
986 if (spc
&& spc
< dollar
-1) {
987 /* There are spaces in unexpected places.
988 * This is probably an id from some other
989 * versioning system. Keep it for now.
994 len
-= dollar
+ 1 - src
;
997 /* it wasn't a "Id$" or "Id:xxxx$" */
1001 /* step 4: substitute */
1002 strbuf_addstr(buf
, "Id: ");
1003 strbuf_add(buf
, sha1_to_hex(sha1
), 40);
1004 strbuf_addstr(buf
, " $");
1006 strbuf_add(buf
, src
, len
);
1012 static enum crlf_action
git_path_check_crlf(struct attr_check_item
*check
)
1014 const char *value
= check
->value
;
1016 if (ATTR_TRUE(value
))
1018 else if (ATTR_FALSE(value
))
1020 else if (ATTR_UNSET(value
))
1022 else if (!strcmp(value
, "input"))
1023 return CRLF_TEXT_INPUT
;
1024 else if (!strcmp(value
, "auto"))
1026 return CRLF_UNDEFINED
;
1029 static enum eol
git_path_check_eol(struct attr_check_item
*check
)
1031 const char *value
= check
->value
;
1033 if (ATTR_UNSET(value
))
1035 else if (!strcmp(value
, "lf"))
1037 else if (!strcmp(value
, "crlf"))
1042 static struct convert_driver
*git_path_check_convert(struct attr_check_item
*check
)
1044 const char *value
= check
->value
;
1045 struct convert_driver
*drv
;
1047 if (ATTR_TRUE(value
) || ATTR_FALSE(value
) || ATTR_UNSET(value
))
1049 for (drv
= user_convert
; drv
; drv
= drv
->next
)
1050 if (!strcmp(value
, drv
->name
))
1055 static int git_path_check_ident(struct attr_check_item
*check
)
1057 const char *value
= check
->value
;
1059 return !!ATTR_TRUE(value
);
1063 struct convert_driver
*drv
;
1064 enum crlf_action attr_action
; /* What attr says */
1065 enum crlf_action crlf_action
; /* When no attr is set, use core.autocrlf */
1069 static void convert_attrs(struct conv_attrs
*ca
, const char *path
)
1071 static struct attr_check
*check
;
1074 check
= attr_check_initl("crlf", "ident", "filter",
1075 "eol", "text", NULL
);
1076 user_convert_tail
= &user_convert
;
1077 git_config(read_convert_config
, NULL
);
1080 if (!git_check_attr(path
, check
)) {
1081 struct attr_check_item
*ccheck
= check
->items
;
1082 ca
->crlf_action
= git_path_check_crlf(ccheck
+ 4);
1083 if (ca
->crlf_action
== CRLF_UNDEFINED
)
1084 ca
->crlf_action
= git_path_check_crlf(ccheck
+ 0);
1085 ca
->attr_action
= ca
->crlf_action
;
1086 ca
->ident
= git_path_check_ident(ccheck
+ 1);
1087 ca
->drv
= git_path_check_convert(ccheck
+ 2);
1088 if (ca
->crlf_action
!= CRLF_BINARY
) {
1089 enum eol eol_attr
= git_path_check_eol(ccheck
+ 3);
1090 if (ca
->crlf_action
== CRLF_AUTO
&& eol_attr
== EOL_LF
)
1091 ca
->crlf_action
= CRLF_AUTO_INPUT
;
1092 else if (ca
->crlf_action
== CRLF_AUTO
&& eol_attr
== EOL_CRLF
)
1093 ca
->crlf_action
= CRLF_AUTO_CRLF
;
1094 else if (eol_attr
== EOL_LF
)
1095 ca
->crlf_action
= CRLF_TEXT_INPUT
;
1096 else if (eol_attr
== EOL_CRLF
)
1097 ca
->crlf_action
= CRLF_TEXT_CRLF
;
1099 ca
->attr_action
= ca
->crlf_action
;
1102 ca
->crlf_action
= CRLF_UNDEFINED
;
1105 if (ca
->crlf_action
== CRLF_TEXT
)
1106 ca
->crlf_action
= text_eol_is_crlf() ? CRLF_TEXT_CRLF
: CRLF_TEXT_INPUT
;
1107 if (ca
->crlf_action
== CRLF_UNDEFINED
&& auto_crlf
== AUTO_CRLF_FALSE
)
1108 ca
->crlf_action
= CRLF_BINARY
;
1109 if (ca
->crlf_action
== CRLF_UNDEFINED
&& auto_crlf
== AUTO_CRLF_TRUE
)
1110 ca
->crlf_action
= CRLF_AUTO_CRLF
;
1111 if (ca
->crlf_action
== CRLF_UNDEFINED
&& auto_crlf
== AUTO_CRLF_INPUT
)
1112 ca
->crlf_action
= CRLF_AUTO_INPUT
;
1115 int would_convert_to_git_filter_fd(const char *path
)
1117 struct conv_attrs ca
;
1119 convert_attrs(&ca
, path
);
1124 * Apply a filter to an fd only if the filter is required to succeed.
1125 * We must die if the filter fails, because the original data before
1126 * filtering is not available.
1128 if (!ca
.drv
->required
)
1131 return apply_filter(path
, NULL
, 0, -1, NULL
, ca
.drv
, CAP_CLEAN
);
1134 const char *get_convert_attr_ascii(const char *path
)
1136 struct conv_attrs ca
;
1138 convert_attrs(&ca
, path
);
1139 switch (ca
.attr_action
) {
1140 case CRLF_UNDEFINED
:
1146 case CRLF_TEXT_INPUT
:
1147 return "text eol=lf";
1148 case CRLF_TEXT_CRLF
:
1149 return "text eol=crlf";
1152 case CRLF_AUTO_CRLF
:
1153 return "text=auto eol=crlf";
1154 case CRLF_AUTO_INPUT
:
1155 return "text=auto eol=lf";
1160 int convert_to_git(const char *path
, const char *src
, size_t len
,
1161 struct strbuf
*dst
, enum safe_crlf checksafe
)
1164 struct conv_attrs ca
;
1166 convert_attrs(&ca
, path
);
1168 ret
|= apply_filter(path
, src
, len
, -1, dst
, ca
.drv
, CAP_CLEAN
);
1169 if (!ret
&& ca
.drv
&& ca
.drv
->required
)
1170 die("%s: clean filter '%s' failed", path
, ca
.drv
->name
);
1176 ret
|= crlf_to_git(path
, src
, len
, dst
, ca
.crlf_action
, checksafe
);
1181 return ret
| ident_to_git(path
, src
, len
, dst
, ca
.ident
);
1184 void convert_to_git_filter_fd(const char *path
, int fd
, struct strbuf
*dst
,
1185 enum safe_crlf checksafe
)
1187 struct conv_attrs ca
;
1188 convert_attrs(&ca
, path
);
1191 assert(ca
.drv
->clean
|| ca
.drv
->process
);
1193 if (!apply_filter(path
, NULL
, 0, fd
, dst
, ca
.drv
, CAP_CLEAN
))
1194 die("%s: clean filter '%s' failed", path
, ca
.drv
->name
);
1196 crlf_to_git(path
, dst
->buf
, dst
->len
, dst
, ca
.crlf_action
, checksafe
);
1197 ident_to_git(path
, dst
->buf
, dst
->len
, dst
, ca
.ident
);
1200 static int convert_to_working_tree_internal(const char *path
, const char *src
,
1201 size_t len
, struct strbuf
*dst
,
1204 int ret
= 0, ret_filter
= 0;
1205 struct conv_attrs ca
;
1207 convert_attrs(&ca
, path
);
1209 ret
|= ident_to_worktree(path
, src
, len
, dst
, ca
.ident
);
1215 * CRLF conversion can be skipped if normalizing, unless there
1216 * is a smudge or process filter (even if the process filter doesn't
1217 * support smudge). The filters might expect CRLFs.
1219 if ((ca
.drv
&& (ca
.drv
->smudge
|| ca
.drv
->process
)) || !normalizing
) {
1220 ret
|= crlf_to_worktree(path
, src
, len
, dst
, ca
.crlf_action
);
1227 ret_filter
= apply_filter(path
, src
, len
, -1, dst
, ca
.drv
, CAP_SMUDGE
);
1228 if (!ret_filter
&& ca
.drv
&& ca
.drv
->required
)
1229 die("%s: smudge filter %s failed", path
, ca
.drv
->name
);
1231 return ret
| ret_filter
;
1234 int convert_to_working_tree(const char *path
, const char *src
, size_t len
, struct strbuf
*dst
)
1236 return convert_to_working_tree_internal(path
, src
, len
, dst
, 0);
1239 int renormalize_buffer(const char *path
, const char *src
, size_t len
, struct strbuf
*dst
)
1241 int ret
= convert_to_working_tree_internal(path
, src
, len
, dst
, 1);
1246 return ret
| convert_to_git(path
, src
, len
, dst
, SAFE_CRLF_RENORMALIZE
);
1249 /*****************************************************************
1251 * Streaming conversion support
1253 *****************************************************************/
1255 typedef int (*filter_fn
)(struct stream_filter
*,
1256 const char *input
, size_t *isize_p
,
1257 char *output
, size_t *osize_p
);
1258 typedef void (*free_fn
)(struct stream_filter
*);
1260 struct stream_filter_vtbl
{
1265 struct stream_filter
{
1266 struct stream_filter_vtbl
*vtbl
;
1269 static int null_filter_fn(struct stream_filter
*filter
,
1270 const char *input
, size_t *isize_p
,
1271 char *output
, size_t *osize_p
)
1276 return 0; /* we do not keep any states */
1278 if (*osize_p
< count
)
1281 memmove(output
, input
, count
);
1288 static void null_free_fn(struct stream_filter
*filter
)
1290 ; /* nothing -- null instances are shared */
1293 static struct stream_filter_vtbl null_vtbl
= {
1298 static struct stream_filter null_filter_singleton
= {
1302 int is_null_stream_filter(struct stream_filter
*filter
)
1304 return filter
== &null_filter_singleton
;
1312 struct lf_to_crlf_filter
{
1313 struct stream_filter filter
;
1314 unsigned has_held
:1;
1318 static int lf_to_crlf_filter_fn(struct stream_filter
*filter
,
1319 const char *input
, size_t *isize_p
,
1320 char *output
, size_t *osize_p
)
1322 size_t count
, o
= 0;
1323 struct lf_to_crlf_filter
*lf_to_crlf
= (struct lf_to_crlf_filter
*)filter
;
1326 * We may be holding onto the CR to see if it is followed by a
1327 * LF, in which case we would need to go to the main loop.
1328 * Otherwise, just emit it to the output stream.
1330 if (lf_to_crlf
->has_held
&& (lf_to_crlf
->held
!= '\r' || !input
)) {
1331 output
[o
++] = lf_to_crlf
->held
;
1332 lf_to_crlf
->has_held
= 0;
1335 /* We are told to drain */
1342 if (count
|| lf_to_crlf
->has_held
) {
1346 if (lf_to_crlf
->has_held
) {
1348 lf_to_crlf
->has_held
= 0;
1351 for (i
= 0; o
< *osize_p
&& i
< count
; i
++) {
1356 } else if (was_cr
) {
1358 * Previous round saw CR and it is not followed
1359 * by a LF; emit the CR before processing the
1360 * current character.
1366 * We may have consumed the last output slot,
1367 * in which case we need to break out of this
1368 * loop; hold the current character before
1371 if (*osize_p
<= o
) {
1372 lf_to_crlf
->has_held
= 1;
1373 lf_to_crlf
->held
= ch
;
1374 continue; /* break but increment i */
1389 if (!lf_to_crlf
->has_held
&& was_cr
) {
1390 lf_to_crlf
->has_held
= 1;
1391 lf_to_crlf
->held
= '\r';
1397 static void lf_to_crlf_free_fn(struct stream_filter
*filter
)
1402 static struct stream_filter_vtbl lf_to_crlf_vtbl
= {
1403 lf_to_crlf_filter_fn
,
1407 static struct stream_filter
*lf_to_crlf_filter(void)
1409 struct lf_to_crlf_filter
*lf_to_crlf
= xcalloc(1, sizeof(*lf_to_crlf
));
1411 lf_to_crlf
->filter
.vtbl
= &lf_to_crlf_vtbl
;
1412 return (struct stream_filter
*)lf_to_crlf
;
1418 #define FILTER_BUFFER 1024
1419 struct cascade_filter
{
1420 struct stream_filter filter
;
1421 struct stream_filter
*one
;
1422 struct stream_filter
*two
;
1423 char buf
[FILTER_BUFFER
];
1427 static int cascade_filter_fn(struct stream_filter
*filter
,
1428 const char *input
, size_t *isize_p
,
1429 char *output
, size_t *osize_p
)
1431 struct cascade_filter
*cas
= (struct cascade_filter
*) filter
;
1433 size_t sz
= *osize_p
;
1434 size_t to_feed
, remaining
;
1437 * input -- (one) --> buf -- (two) --> output
1439 while (filled
< sz
) {
1440 remaining
= sz
- filled
;
1442 /* do we already have something to feed two with? */
1443 if (cas
->ptr
< cas
->end
) {
1444 to_feed
= cas
->end
- cas
->ptr
;
1445 if (stream_filter(cas
->two
,
1446 cas
->buf
+ cas
->ptr
, &to_feed
,
1447 output
+ filled
, &remaining
))
1449 cas
->ptr
+= (cas
->end
- cas
->ptr
) - to_feed
;
1450 filled
= sz
- remaining
;
1454 /* feed one from upstream and have it emit into our buffer */
1455 to_feed
= input
? *isize_p
: 0;
1456 if (input
&& !to_feed
)
1458 remaining
= sizeof(cas
->buf
);
1459 if (stream_filter(cas
->one
,
1461 cas
->buf
, &remaining
))
1463 cas
->end
= sizeof(cas
->buf
) - remaining
;
1466 size_t fed
= *isize_p
- to_feed
;
1471 /* do we know that we drained one completely? */
1472 if (input
|| cas
->end
)
1475 /* tell two to drain; we have nothing more to give it */
1477 remaining
= sz
- filled
;
1478 if (stream_filter(cas
->two
,
1480 output
+ filled
, &remaining
))
1482 if (remaining
== (sz
- filled
))
1483 break; /* completely drained two */
1484 filled
= sz
- remaining
;
1490 static void cascade_free_fn(struct stream_filter
*filter
)
1492 struct cascade_filter
*cas
= (struct cascade_filter
*)filter
;
1493 free_stream_filter(cas
->one
);
1494 free_stream_filter(cas
->two
);
1498 static struct stream_filter_vtbl cascade_vtbl
= {
1503 static struct stream_filter
*cascade_filter(struct stream_filter
*one
,
1504 struct stream_filter
*two
)
1506 struct cascade_filter
*cascade
;
1508 if (!one
|| is_null_stream_filter(one
))
1510 if (!two
|| is_null_stream_filter(two
))
1513 cascade
= xmalloc(sizeof(*cascade
));
1516 cascade
->end
= cascade
->ptr
= 0;
1517 cascade
->filter
.vtbl
= &cascade_vtbl
;
1518 return (struct stream_filter
*)cascade
;
1524 #define IDENT_DRAINING (-1)
1525 #define IDENT_SKIPPING (-2)
1526 struct ident_filter
{
1527 struct stream_filter filter
;
1530 char ident
[45]; /* ": x40 $" */
1533 static int is_foreign_ident(const char *str
)
1537 if (!skip_prefix(str
, "$Id: ", &str
))
1539 for (i
= 0; str
[i
]; i
++) {
1540 if (isspace(str
[i
]) && str
[i
+1] != '$')
1546 static void ident_drain(struct ident_filter
*ident
, char **output_p
, size_t *osize_p
)
1548 size_t to_drain
= ident
->left
.len
;
1550 if (*osize_p
< to_drain
)
1551 to_drain
= *osize_p
;
1553 memcpy(*output_p
, ident
->left
.buf
, to_drain
);
1554 strbuf_remove(&ident
->left
, 0, to_drain
);
1555 *output_p
+= to_drain
;
1556 *osize_p
-= to_drain
;
1558 if (!ident
->left
.len
)
1562 static int ident_filter_fn(struct stream_filter
*filter
,
1563 const char *input
, size_t *isize_p
,
1564 char *output
, size_t *osize_p
)
1566 struct ident_filter
*ident
= (struct ident_filter
*)filter
;
1567 static const char head
[] = "$Id";
1570 /* drain upon eof */
1571 switch (ident
->state
) {
1573 strbuf_add(&ident
->left
, head
, ident
->state
);
1574 case IDENT_SKIPPING
:
1576 case IDENT_DRAINING
:
1577 ident_drain(ident
, &output
, osize_p
);
1582 while (*isize_p
|| (ident
->state
== IDENT_DRAINING
)) {
1585 if (ident
->state
== IDENT_DRAINING
) {
1586 ident_drain(ident
, &output
, osize_p
);
1595 if (ident
->state
== IDENT_SKIPPING
) {
1597 * Skipping until '$' or LF, but keeping them
1598 * in case it is a foreign ident.
1600 strbuf_addch(&ident
->left
, ch
);
1601 if (ch
!= '\n' && ch
!= '$')
1603 if (ch
== '$' && !is_foreign_ident(ident
->left
.buf
)) {
1604 strbuf_setlen(&ident
->left
, sizeof(head
) - 1);
1605 strbuf_addstr(&ident
->left
, ident
->ident
);
1607 ident
->state
= IDENT_DRAINING
;
1611 if (ident
->state
< sizeof(head
) &&
1612 head
[ident
->state
] == ch
) {
1618 strbuf_add(&ident
->left
, head
, ident
->state
);
1619 if (ident
->state
== sizeof(head
) - 1) {
1620 if (ch
!= ':' && ch
!= '$') {
1621 strbuf_addch(&ident
->left
, ch
);
1627 strbuf_addch(&ident
->left
, ch
);
1628 ident
->state
= IDENT_SKIPPING
;
1630 strbuf_addstr(&ident
->left
, ident
->ident
);
1631 ident
->state
= IDENT_DRAINING
;
1636 strbuf_addch(&ident
->left
, ch
);
1637 ident
->state
= IDENT_DRAINING
;
1642 static void ident_free_fn(struct stream_filter
*filter
)
1644 struct ident_filter
*ident
= (struct ident_filter
*)filter
;
1645 strbuf_release(&ident
->left
);
1649 static struct stream_filter_vtbl ident_vtbl
= {
1654 static struct stream_filter
*ident_filter(const unsigned char *sha1
)
1656 struct ident_filter
*ident
= xmalloc(sizeof(*ident
));
1658 xsnprintf(ident
->ident
, sizeof(ident
->ident
),
1659 ": %s $", sha1_to_hex(sha1
));
1660 strbuf_init(&ident
->left
, 0);
1661 ident
->filter
.vtbl
= &ident_vtbl
;
1663 return (struct stream_filter
*)ident
;
1667 * Return an appropriately constructed filter for the path, or NULL if
1668 * the contents cannot be filtered without reading the whole thing
1671 * Note that you would be crazy to set CRLF, smuge/clean or ident to a
1672 * large binary blob you would want us not to slurp into the memory!
1674 struct stream_filter
*get_stream_filter(const char *path
, const unsigned char *sha1
)
1676 struct conv_attrs ca
;
1677 struct stream_filter
*filter
= NULL
;
1679 convert_attrs(&ca
, path
);
1680 if (ca
.drv
&& (ca
.drv
->process
|| ca
.drv
->smudge
|| ca
.drv
->clean
))
1683 if (ca
.crlf_action
== CRLF_AUTO
|| ca
.crlf_action
== CRLF_AUTO_CRLF
)
1687 filter
= ident_filter(sha1
);
1689 if (output_eol(ca
.crlf_action
) == EOL_CRLF
)
1690 filter
= cascade_filter(filter
, lf_to_crlf_filter());
1692 filter
= cascade_filter(filter
, &null_filter_singleton
);
1697 void free_stream_filter(struct stream_filter
*filter
)
1699 filter
->vtbl
->free(filter
);
1702 int stream_filter(struct stream_filter
*filter
,
1703 const char *input
, size_t *isize_p
,
1704 char *output
, size_t *osize_p
)
1706 return filter
->vtbl
->filter(filter
, input
, isize_p
, output
, osize_p
);