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.
16 /* Stat bits: When BIN is set, the txt bits are unset */
17 #define CONVERT_STAT_BITS_TXT_LF 0x1
18 #define CONVERT_STAT_BITS_TXT_CRLF 0x2
19 #define CONVERT_STAT_BITS_BIN 0x4
31 /* NUL, CR, LF and CRLF counts */
32 unsigned nul
, cr
, lf
, crlf
;
34 /* These are just approximations! */
35 unsigned printable
, nonprintable
;
38 static void gather_stats(const char *buf
, unsigned long size
, struct text_stat
*stats
)
42 memset(stats
, 0, sizeof(*stats
));
44 for (i
= 0; i
< size
; i
++) {
45 unsigned char c
= buf
[i
];
48 if (i
+1 < size
&& buf
[i
+1] == '\n')
58 stats
->nonprintable
++;
61 /* BS, HT, ESC and FF */
62 case '\b': case '\t': case '\033': case '\014':
69 stats
->nonprintable
++;
76 /* If file ends with EOF then don't count this EOF as non-printable. */
77 if (size
>= 1 && buf
[size
-1] == '\032')
78 stats
->nonprintable
--;
82 * The same heuristics as diff.c::mmfile_is_binary()
83 * We treat files with bare CR as binary
85 static int convert_is_binary(unsigned long size
, const struct text_stat
*stats
)
87 if (stats
->cr
!= stats
->crlf
)
91 if ((stats
->printable
>> 7) < stats
->nonprintable
)
96 static unsigned int gather_convert_stats(const char *data
, unsigned long size
)
98 struct text_stat stats
;
101 gather_stats(data
, size
, &stats
);
102 if (convert_is_binary(size
, &stats
))
103 return CONVERT_STAT_BITS_BIN
;
104 else if (stats
.crlf
&& stats
.crlf
== stats
.lf
)
105 return CONVERT_STAT_BITS_TXT_CRLF
;
106 else if (stats
.crlf
&& stats
.lf
)
107 return CONVERT_STAT_BITS_TXT_CRLF
| CONVERT_STAT_BITS_TXT_LF
;
109 return CONVERT_STAT_BITS_TXT_LF
;
114 static const char *gather_convert_stats_ascii(const char *data
, unsigned long size
)
116 unsigned int convert_stats
= gather_convert_stats(data
, size
);
118 if (convert_stats
& CONVERT_STAT_BITS_BIN
)
120 switch (convert_stats
) {
121 case CONVERT_STAT_BITS_TXT_LF
:
123 case CONVERT_STAT_BITS_TXT_CRLF
:
125 case CONVERT_STAT_BITS_TXT_LF
| CONVERT_STAT_BITS_TXT_CRLF
:
132 const char *get_cached_convert_stats_ascii(const char *path
)
136 void *data
= read_blob_data_from_cache(path
, &sz
);
137 ret
= gather_convert_stats_ascii(data
, sz
);
142 const char *get_wt_convert_stats_ascii(const char *path
)
144 const char *ret
= "";
145 struct strbuf sb
= STRBUF_INIT
;
146 if (strbuf_read_file(&sb
, path
, 0) >= 0)
147 ret
= gather_convert_stats_ascii(sb
.buf
, sb
.len
);
152 static enum eol
output_eol(enum crlf_action crlf_action
)
154 switch (crlf_action
) {
167 if (auto_crlf
== AUTO_CRLF_TRUE
)
169 else if (auto_crlf
== AUTO_CRLF_INPUT
)
171 else if (core_eol
== EOL_UNSET
)
177 static void check_safe_crlf(const char *path
, enum crlf_action crlf_action
,
178 struct text_stat
*stats
, enum safe_crlf checksafe
)
183 if (output_eol(crlf_action
) == EOL_LF
) {
185 * CRLFs would not be restored by checkout:
186 * check if we'd remove CRLFs
189 if (checksafe
== SAFE_CRLF_WARN
)
190 warning("CRLF will be replaced by LF in %s.\nThe file will have its original line endings in your working directory.", path
);
191 else /* i.e. SAFE_CRLF_FAIL */
192 die("CRLF would be replaced by LF in %s.", path
);
194 } else if (output_eol(crlf_action
) == EOL_CRLF
) {
196 * CRLFs would be added by checkout:
197 * check if we have "naked" LFs
199 if (stats
->lf
!= stats
->crlf
) {
200 if (checksafe
== SAFE_CRLF_WARN
)
201 warning("LF will be replaced by CRLF in %s.\nThe file will have its original line endings in your working directory.", path
);
202 else /* i.e. SAFE_CRLF_FAIL */
203 die("LF would be replaced by CRLF in %s", path
);
208 static int has_cr_in_index(const char *path
)
214 data
= read_blob_data_from_cache(path
, &sz
);
217 has_cr
= memchr(data
, '\r', sz
) != NULL
;
222 static int crlf_to_git(const char *path
, const char *src
, size_t len
,
224 enum crlf_action crlf_action
, enum safe_crlf checksafe
)
226 struct text_stat stats
;
229 if (crlf_action
== CRLF_BINARY
||
230 (crlf_action
== CRLF_GUESS
&& auto_crlf
== AUTO_CRLF_FALSE
) ||
235 * If we are doing a dry-run and have no source buffer, there is
236 * nothing to analyze; we must assume we would convert.
241 gather_stats(src
, len
, &stats
);
243 if (crlf_action
== CRLF_AUTO
|| crlf_action
== CRLF_GUESS
) {
244 if (convert_is_binary(len
, &stats
))
247 if (crlf_action
== CRLF_GUESS
) {
249 * If the file in the index has any CR in it, do not convert.
250 * This is the new safer autocrlf handling.
252 if (has_cr_in_index(path
))
257 check_safe_crlf(path
, crlf_action
, &stats
, checksafe
);
259 /* Optimization: No CR? Nothing to convert, regardless. */
264 * At this point all of our source analysis is done, and we are sure we
265 * would convert. If we are in dry-run mode, we can give an answer.
270 /* only grow if not in place */
271 if (strbuf_avail(buf
) + buf
->len
< len
)
272 strbuf_grow(buf
, len
- buf
->len
);
274 if (crlf_action
== CRLF_AUTO
|| crlf_action
== CRLF_GUESS
) {
276 * If we guessed, we already know we rejected a file with
277 * lone CR, and we can strip a CR without looking at what
281 unsigned char c
= *src
++;
287 unsigned char c
= *src
++;
288 if (! (c
== '\r' && (1 < len
&& *src
== '\n')))
292 strbuf_setlen(buf
, dst
- buf
->buf
);
296 static int crlf_to_worktree(const char *path
, const char *src
, size_t len
,
297 struct strbuf
*buf
, enum crlf_action crlf_action
)
299 char *to_free
= NULL
;
300 struct text_stat stats
;
302 if (!len
|| output_eol(crlf_action
) != EOL_CRLF
)
305 gather_stats(src
, len
, &stats
);
307 /* No LF? Nothing to convert, regardless. */
311 /* Was it already in CRLF format? */
312 if (stats
.lf
== stats
.crlf
)
315 if (crlf_action
== CRLF_AUTO
|| crlf_action
== CRLF_GUESS
) {
316 if (crlf_action
== CRLF_GUESS
) {
317 /* If we have any CR or CRLF line endings, we do not touch it */
318 /* This is the new safer autocrlf-handling */
319 if (stats
.cr
> 0 || stats
.crlf
> 0)
323 if (convert_is_binary(len
, &stats
))
327 /* are we "faking" in place editing ? */
329 to_free
= strbuf_detach(buf
, NULL
);
331 strbuf_grow(buf
, len
+ stats
.lf
- stats
.crlf
);
333 const char *nl
= memchr(src
, '\n', len
);
336 if (nl
> src
&& nl
[-1] == '\r') {
337 strbuf_add(buf
, src
, nl
+ 1 - src
);
339 strbuf_add(buf
, src
, nl
- src
);
340 strbuf_addstr(buf
, "\r\n");
345 strbuf_add(buf
, src
, len
);
351 struct filter_params
{
359 static int filter_buffer_or_fd(int in
, int out
, void *data
)
362 * Spawn cmd and feed the buffer contents through its stdin.
364 struct child_process child_process
= CHILD_PROCESS_INIT
;
365 struct filter_params
*params
= (struct filter_params
*)data
;
366 int write_err
, status
;
367 const char *argv
[] = { NULL
, NULL
};
369 /* apply % substitution to cmd */
370 struct strbuf cmd
= STRBUF_INIT
;
371 struct strbuf path
= STRBUF_INIT
;
372 struct strbuf_expand_dict_entry dict
[] = {
377 /* quote the path to preserve spaces, etc. */
378 sq_quote_buf(&path
, params
->path
);
379 dict
[0].value
= path
.buf
;
381 /* expand all %f with the quoted path */
382 strbuf_expand(&cmd
, params
->cmd
, strbuf_expand_dict_cb
, &dict
);
383 strbuf_release(&path
);
387 child_process
.argv
= argv
;
388 child_process
.use_shell
= 1;
389 child_process
.in
= -1;
390 child_process
.out
= out
;
392 if (start_command(&child_process
))
393 return error("cannot fork to run external filter %s", params
->cmd
);
395 sigchain_push(SIGPIPE
, SIG_IGN
);
398 write_err
= (write_in_full(child_process
.in
,
399 params
->src
, params
->size
) < 0);
403 write_err
= copy_fd(params
->fd
, child_process
.in
);
404 if (write_err
== COPY_WRITE_ERROR
&& errno
== EPIPE
)
408 if (close(child_process
.in
))
411 error("cannot feed the input to external filter %s", params
->cmd
);
413 sigchain_pop(SIGPIPE
);
415 status
= finish_command(&child_process
);
417 error("external filter %s failed %d", params
->cmd
, status
);
419 strbuf_release(&cmd
);
420 return (write_err
|| status
);
423 static int apply_filter(const char *path
, const char *src
, size_t len
, int fd
,
424 struct strbuf
*dst
, const char *cmd
)
427 * Create a pipeline to have the command filter the buffer's
430 * (child --> cmd) --> us
433 struct strbuf nbuf
= STRBUF_INIT
;
435 struct filter_params params
;
443 memset(&async
, 0, sizeof(async
));
444 async
.proc
= filter_buffer_or_fd
;
445 async
.data
= ¶ms
;
454 if (start_async(&async
))
455 return 0; /* error was already reported */
457 if (strbuf_read(&nbuf
, async
.out
, len
) < 0) {
458 error("read from external filter %s failed", cmd
);
461 if (close(async
.out
)) {
462 error("read from external filter %s failed", cmd
);
465 if (finish_async(&async
)) {
466 error("external filter %s failed", cmd
);
471 strbuf_swap(dst
, &nbuf
);
473 strbuf_release(&nbuf
);
477 static struct convert_driver
{
479 struct convert_driver
*next
;
483 } *user_convert
, **user_convert_tail
;
485 static int read_convert_config(const char *var
, const char *value
, void *cb
)
487 const char *key
, *name
;
489 struct convert_driver
*drv
;
492 * External conversion drivers are configured using
493 * "filter.<name>.variable".
495 if (parse_config_key(var
, "filter", &name
, &namelen
, &key
) < 0 || !name
)
497 for (drv
= user_convert
; drv
; drv
= drv
->next
)
498 if (!strncmp(drv
->name
, name
, namelen
) && !drv
->name
[namelen
])
501 drv
= xcalloc(1, sizeof(struct convert_driver
));
502 drv
->name
= xmemdupz(name
, namelen
);
503 *user_convert_tail
= drv
;
504 user_convert_tail
= &(drv
->next
);
508 * filter.<name>.smudge and filter.<name>.clean specifies
513 * The command-line will not be interpolated in any way.
516 if (!strcmp("smudge", key
))
517 return git_config_string(&drv
->smudge
, var
, value
);
519 if (!strcmp("clean", key
))
520 return git_config_string(&drv
->clean
, var
, value
);
522 if (!strcmp("required", key
)) {
523 drv
->required
= git_config_bool(var
, value
);
530 static int count_ident(const char *cp
, unsigned long size
)
533 * "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$"
545 if (memcmp("Id", cp
, 2))
556 * "$Id: ... "; scan up to the closing dollar sign and discard.
572 static int ident_to_git(const char *path
, const char *src
, size_t len
,
573 struct strbuf
*buf
, int ident
)
577 if (!ident
|| (src
&& !count_ident(src
, len
)))
583 /* only grow if not in place */
584 if (strbuf_avail(buf
) + buf
->len
< len
)
585 strbuf_grow(buf
, len
- buf
->len
);
588 dollar
= memchr(src
, '$', len
);
591 memmove(dst
, src
, dollar
+ 1 - src
);
592 dst
+= dollar
+ 1 - src
;
593 len
-= dollar
+ 1 - src
;
596 if (len
> 3 && !memcmp(src
, "Id:", 3)) {
597 dollar
= memchr(src
+ 3, '$', len
- 3);
600 if (memchr(src
+ 3, '\n', dollar
- src
- 3)) {
601 /* Line break before the next dollar. */
605 memcpy(dst
, "Id$", 3);
607 len
-= dollar
+ 1 - src
;
611 memmove(dst
, src
, len
);
612 strbuf_setlen(buf
, dst
+ len
- buf
->buf
);
616 static int ident_to_worktree(const char *path
, const char *src
, size_t len
,
617 struct strbuf
*buf
, int ident
)
619 unsigned char sha1
[20];
620 char *to_free
= NULL
, *dollar
, *spc
;
626 cnt
= count_ident(src
, len
);
630 /* are we "faking" in place editing ? */
632 to_free
= strbuf_detach(buf
, NULL
);
633 hash_sha1_file(src
, len
, "blob", sha1
);
635 strbuf_grow(buf
, len
+ cnt
* 43);
637 /* step 1: run to the next '$' */
638 dollar
= memchr(src
, '$', len
);
641 strbuf_add(buf
, src
, dollar
+ 1 - src
);
642 len
-= dollar
+ 1 - src
;
645 /* step 2: does it looks like a bit like Id:xxx$ or Id$ ? */
646 if (len
< 3 || memcmp("Id", src
, 2))
649 /* step 3: skip over Id$ or Id:xxxxx$ */
653 } else if (src
[2] == ':') {
655 * It's possible that an expanded Id has crept its way into the
656 * repository, we cope with that by stripping the expansion out.
657 * This is probably not a good idea, since it will cause changes
658 * on checkout, which won't go away by stash, but let's keep it
661 dollar
= memchr(src
+ 3, '$', len
- 3);
663 /* incomplete keyword, no more '$', so just quit the loop */
667 if (memchr(src
+ 3, '\n', dollar
- src
- 3)) {
668 /* Line break before the next dollar. */
672 spc
= memchr(src
+ 4, ' ', dollar
- src
- 4);
673 if (spc
&& spc
< dollar
-1) {
674 /* There are spaces in unexpected places.
675 * This is probably an id from some other
676 * versioning system. Keep it for now.
681 len
-= dollar
+ 1 - src
;
684 /* it wasn't a "Id$" or "Id:xxxx$" */
688 /* step 4: substitute */
689 strbuf_addstr(buf
, "Id: ");
690 strbuf_add(buf
, sha1_to_hex(sha1
), 40);
691 strbuf_addstr(buf
, " $");
693 strbuf_add(buf
, src
, len
);
699 static enum crlf_action
git_path_check_crlf(const char *path
, struct git_attr_check
*check
)
701 const char *value
= check
->value
;
703 if (ATTR_TRUE(value
))
705 else if (ATTR_FALSE(value
))
707 else if (ATTR_UNSET(value
))
709 else if (!strcmp(value
, "input"))
711 else if (!strcmp(value
, "auto"))
716 static enum eol
git_path_check_eol(const char *path
, struct git_attr_check
*check
)
718 const char *value
= check
->value
;
720 if (ATTR_UNSET(value
))
722 else if (!strcmp(value
, "lf"))
724 else if (!strcmp(value
, "crlf"))
729 static struct convert_driver
*git_path_check_convert(const char *path
,
730 struct git_attr_check
*check
)
732 const char *value
= check
->value
;
733 struct convert_driver
*drv
;
735 if (ATTR_TRUE(value
) || ATTR_FALSE(value
) || ATTR_UNSET(value
))
737 for (drv
= user_convert
; drv
; drv
= drv
->next
)
738 if (!strcmp(value
, drv
->name
))
743 static int git_path_check_ident(const char *path
, struct git_attr_check
*check
)
745 const char *value
= check
->value
;
747 return !!ATTR_TRUE(value
);
750 static enum crlf_action
input_crlf_action(enum crlf_action text_attr
, enum eol eol_attr
)
752 if (text_attr
== CRLF_BINARY
)
754 if (eol_attr
== EOL_LF
)
756 if (eol_attr
== EOL_CRLF
)
762 struct convert_driver
*drv
;
763 enum crlf_action crlf_action
;
768 static const char *conv_attr_name
[] = {
769 "crlf", "ident", "filter", "eol", "text",
771 #define NUM_CONV_ATTRS ARRAY_SIZE(conv_attr_name)
773 static void convert_attrs(struct conv_attrs
*ca
, const char *path
)
776 static struct git_attr_check ccheck
[NUM_CONV_ATTRS
];
778 if (!ccheck
[0].attr
) {
779 for (i
= 0; i
< NUM_CONV_ATTRS
; i
++)
780 ccheck
[i
].attr
= git_attr(conv_attr_name
[i
]);
781 user_convert_tail
= &user_convert
;
782 git_config(read_convert_config
, NULL
);
785 if (!git_check_attr(path
, NUM_CONV_ATTRS
, ccheck
)) {
786 ca
->crlf_action
= git_path_check_crlf(path
, ccheck
+ 4);
787 if (ca
->crlf_action
== CRLF_GUESS
)
788 ca
->crlf_action
= git_path_check_crlf(path
, ccheck
+ 0);
789 ca
->ident
= git_path_check_ident(path
, ccheck
+ 1);
790 ca
->drv
= git_path_check_convert(path
, ccheck
+ 2);
791 ca
->eol_attr
= git_path_check_eol(path
, ccheck
+ 3);
794 ca
->crlf_action
= CRLF_GUESS
;
795 ca
->eol_attr
= EOL_UNSET
;
800 int would_convert_to_git_filter_fd(const char *path
)
802 struct conv_attrs ca
;
804 convert_attrs(&ca
, path
);
809 * Apply a filter to an fd only if the filter is required to succeed.
810 * We must die if the filter fails, because the original data before
811 * filtering is not available.
813 if (!ca
.drv
->required
)
816 return apply_filter(path
, NULL
, 0, -1, NULL
, ca
.drv
->clean
);
819 const char *get_convert_attr_ascii(const char *path
)
821 struct conv_attrs ca
;
822 enum crlf_action crlf_action
;
824 convert_attrs(&ca
, path
);
825 crlf_action
= input_crlf_action(ca
.crlf_action
, ca
.eol_attr
);
826 switch (crlf_action
) {
834 return "text eol=lf";
836 return "text=auto eol=crlf";
843 int convert_to_git(const char *path
, const char *src
, size_t len
,
844 struct strbuf
*dst
, enum safe_crlf checksafe
)
847 const char *filter
= NULL
;
849 struct conv_attrs ca
;
851 convert_attrs(&ca
, path
);
853 filter
= ca
.drv
->clean
;
854 required
= ca
.drv
->required
;
857 ret
|= apply_filter(path
, src
, len
, -1, dst
, filter
);
858 if (!ret
&& required
)
859 die("%s: clean filter '%s' failed", path
, ca
.drv
->name
);
865 ca
.crlf_action
= input_crlf_action(ca
.crlf_action
, ca
.eol_attr
);
866 ret
|= crlf_to_git(path
, src
, len
, dst
, ca
.crlf_action
, checksafe
);
871 return ret
| ident_to_git(path
, src
, len
, dst
, ca
.ident
);
874 void convert_to_git_filter_fd(const char *path
, int fd
, struct strbuf
*dst
,
875 enum safe_crlf checksafe
)
877 struct conv_attrs ca
;
878 convert_attrs(&ca
, path
);
881 assert(ca
.drv
->clean
);
883 if (!apply_filter(path
, NULL
, 0, fd
, dst
, ca
.drv
->clean
))
884 die("%s: clean filter '%s' failed", path
, ca
.drv
->name
);
886 ca
.crlf_action
= input_crlf_action(ca
.crlf_action
, ca
.eol_attr
);
887 crlf_to_git(path
, dst
->buf
, dst
->len
, dst
, ca
.crlf_action
, checksafe
);
888 ident_to_git(path
, dst
->buf
, dst
->len
, dst
, ca
.ident
);
891 static int convert_to_working_tree_internal(const char *path
, const char *src
,
892 size_t len
, struct strbuf
*dst
,
895 int ret
= 0, ret_filter
= 0;
896 const char *filter
= NULL
;
898 struct conv_attrs ca
;
900 convert_attrs(&ca
, path
);
902 filter
= ca
.drv
->smudge
;
903 required
= ca
.drv
->required
;
906 ret
|= ident_to_worktree(path
, src
, len
, dst
, ca
.ident
);
912 * CRLF conversion can be skipped if normalizing, unless there
913 * is a smudge filter. The filter might expect CRLFs.
915 if (filter
|| !normalizing
) {
916 ca
.crlf_action
= input_crlf_action(ca
.crlf_action
, ca
.eol_attr
);
917 ret
|= crlf_to_worktree(path
, src
, len
, dst
, ca
.crlf_action
);
924 ret_filter
= apply_filter(path
, src
, len
, -1, dst
, filter
);
925 if (!ret_filter
&& required
)
926 die("%s: smudge filter %s failed", path
, ca
.drv
->name
);
928 return ret
| ret_filter
;
931 int convert_to_working_tree(const char *path
, const char *src
, size_t len
, struct strbuf
*dst
)
933 return convert_to_working_tree_internal(path
, src
, len
, dst
, 0);
936 int renormalize_buffer(const char *path
, const char *src
, size_t len
, struct strbuf
*dst
)
938 int ret
= convert_to_working_tree_internal(path
, src
, len
, dst
, 1);
943 return ret
| convert_to_git(path
, src
, len
, dst
, SAFE_CRLF_FALSE
);
946 /*****************************************************************
948 * Streaming conversion support
950 *****************************************************************/
952 typedef int (*filter_fn
)(struct stream_filter
*,
953 const char *input
, size_t *isize_p
,
954 char *output
, size_t *osize_p
);
955 typedef void (*free_fn
)(struct stream_filter
*);
957 struct stream_filter_vtbl
{
962 struct stream_filter
{
963 struct stream_filter_vtbl
*vtbl
;
966 static int null_filter_fn(struct stream_filter
*filter
,
967 const char *input
, size_t *isize_p
,
968 char *output
, size_t *osize_p
)
973 return 0; /* we do not keep any states */
975 if (*osize_p
< count
)
978 memmove(output
, input
, count
);
985 static void null_free_fn(struct stream_filter
*filter
)
987 ; /* nothing -- null instances are shared */
990 static struct stream_filter_vtbl null_vtbl
= {
995 static struct stream_filter null_filter_singleton
= {
999 int is_null_stream_filter(struct stream_filter
*filter
)
1001 return filter
== &null_filter_singleton
;
1009 struct lf_to_crlf_filter
{
1010 struct stream_filter filter
;
1011 unsigned has_held
:1;
1015 static int lf_to_crlf_filter_fn(struct stream_filter
*filter
,
1016 const char *input
, size_t *isize_p
,
1017 char *output
, size_t *osize_p
)
1019 size_t count
, o
= 0;
1020 struct lf_to_crlf_filter
*lf_to_crlf
= (struct lf_to_crlf_filter
*)filter
;
1023 * We may be holding onto the CR to see if it is followed by a
1024 * LF, in which case we would need to go to the main loop.
1025 * Otherwise, just emit it to the output stream.
1027 if (lf_to_crlf
->has_held
&& (lf_to_crlf
->held
!= '\r' || !input
)) {
1028 output
[o
++] = lf_to_crlf
->held
;
1029 lf_to_crlf
->has_held
= 0;
1032 /* We are told to drain */
1039 if (count
|| lf_to_crlf
->has_held
) {
1043 if (lf_to_crlf
->has_held
) {
1045 lf_to_crlf
->has_held
= 0;
1048 for (i
= 0; o
< *osize_p
&& i
< count
; i
++) {
1053 } else if (was_cr
) {
1055 * Previous round saw CR and it is not followed
1056 * by a LF; emit the CR before processing the
1057 * current character.
1063 * We may have consumed the last output slot,
1064 * in which case we need to break out of this
1065 * loop; hold the current character before
1068 if (*osize_p
<= o
) {
1069 lf_to_crlf
->has_held
= 1;
1070 lf_to_crlf
->held
= ch
;
1071 continue; /* break but increment i */
1086 if (!lf_to_crlf
->has_held
&& was_cr
) {
1087 lf_to_crlf
->has_held
= 1;
1088 lf_to_crlf
->held
= '\r';
1094 static void lf_to_crlf_free_fn(struct stream_filter
*filter
)
1099 static struct stream_filter_vtbl lf_to_crlf_vtbl
= {
1100 lf_to_crlf_filter_fn
,
1104 static struct stream_filter
*lf_to_crlf_filter(void)
1106 struct lf_to_crlf_filter
*lf_to_crlf
= xcalloc(1, sizeof(*lf_to_crlf
));
1108 lf_to_crlf
->filter
.vtbl
= &lf_to_crlf_vtbl
;
1109 return (struct stream_filter
*)lf_to_crlf
;
1115 #define FILTER_BUFFER 1024
1116 struct cascade_filter
{
1117 struct stream_filter filter
;
1118 struct stream_filter
*one
;
1119 struct stream_filter
*two
;
1120 char buf
[FILTER_BUFFER
];
1124 static int cascade_filter_fn(struct stream_filter
*filter
,
1125 const char *input
, size_t *isize_p
,
1126 char *output
, size_t *osize_p
)
1128 struct cascade_filter
*cas
= (struct cascade_filter
*) filter
;
1130 size_t sz
= *osize_p
;
1131 size_t to_feed
, remaining
;
1134 * input -- (one) --> buf -- (two) --> output
1136 while (filled
< sz
) {
1137 remaining
= sz
- filled
;
1139 /* do we already have something to feed two with? */
1140 if (cas
->ptr
< cas
->end
) {
1141 to_feed
= cas
->end
- cas
->ptr
;
1142 if (stream_filter(cas
->two
,
1143 cas
->buf
+ cas
->ptr
, &to_feed
,
1144 output
+ filled
, &remaining
))
1146 cas
->ptr
+= (cas
->end
- cas
->ptr
) - to_feed
;
1147 filled
= sz
- remaining
;
1151 /* feed one from upstream and have it emit into our buffer */
1152 to_feed
= input
? *isize_p
: 0;
1153 if (input
&& !to_feed
)
1155 remaining
= sizeof(cas
->buf
);
1156 if (stream_filter(cas
->one
,
1158 cas
->buf
, &remaining
))
1160 cas
->end
= sizeof(cas
->buf
) - remaining
;
1163 size_t fed
= *isize_p
- to_feed
;
1168 /* do we know that we drained one completely? */
1169 if (input
|| cas
->end
)
1172 /* tell two to drain; we have nothing more to give it */
1174 remaining
= sz
- filled
;
1175 if (stream_filter(cas
->two
,
1177 output
+ filled
, &remaining
))
1179 if (remaining
== (sz
- filled
))
1180 break; /* completely drained two */
1181 filled
= sz
- remaining
;
1187 static void cascade_free_fn(struct stream_filter
*filter
)
1189 struct cascade_filter
*cas
= (struct cascade_filter
*)filter
;
1190 free_stream_filter(cas
->one
);
1191 free_stream_filter(cas
->two
);
1195 static struct stream_filter_vtbl cascade_vtbl
= {
1200 static struct stream_filter
*cascade_filter(struct stream_filter
*one
,
1201 struct stream_filter
*two
)
1203 struct cascade_filter
*cascade
;
1205 if (!one
|| is_null_stream_filter(one
))
1207 if (!two
|| is_null_stream_filter(two
))
1210 cascade
= xmalloc(sizeof(*cascade
));
1213 cascade
->end
= cascade
->ptr
= 0;
1214 cascade
->filter
.vtbl
= &cascade_vtbl
;
1215 return (struct stream_filter
*)cascade
;
1221 #define IDENT_DRAINING (-1)
1222 #define IDENT_SKIPPING (-2)
1223 struct ident_filter
{
1224 struct stream_filter filter
;
1227 char ident
[45]; /* ": x40 $" */
1230 static int is_foreign_ident(const char *str
)
1234 if (!skip_prefix(str
, "$Id: ", &str
))
1236 for (i
= 0; str
[i
]; i
++) {
1237 if (isspace(str
[i
]) && str
[i
+1] != '$')
1243 static void ident_drain(struct ident_filter
*ident
, char **output_p
, size_t *osize_p
)
1245 size_t to_drain
= ident
->left
.len
;
1247 if (*osize_p
< to_drain
)
1248 to_drain
= *osize_p
;
1250 memcpy(*output_p
, ident
->left
.buf
, to_drain
);
1251 strbuf_remove(&ident
->left
, 0, to_drain
);
1252 *output_p
+= to_drain
;
1253 *osize_p
-= to_drain
;
1255 if (!ident
->left
.len
)
1259 static int ident_filter_fn(struct stream_filter
*filter
,
1260 const char *input
, size_t *isize_p
,
1261 char *output
, size_t *osize_p
)
1263 struct ident_filter
*ident
= (struct ident_filter
*)filter
;
1264 static const char head
[] = "$Id";
1267 /* drain upon eof */
1268 switch (ident
->state
) {
1270 strbuf_add(&ident
->left
, head
, ident
->state
);
1271 case IDENT_SKIPPING
:
1273 case IDENT_DRAINING
:
1274 ident_drain(ident
, &output
, osize_p
);
1279 while (*isize_p
|| (ident
->state
== IDENT_DRAINING
)) {
1282 if (ident
->state
== IDENT_DRAINING
) {
1283 ident_drain(ident
, &output
, osize_p
);
1292 if (ident
->state
== IDENT_SKIPPING
) {
1294 * Skipping until '$' or LF, but keeping them
1295 * in case it is a foreign ident.
1297 strbuf_addch(&ident
->left
, ch
);
1298 if (ch
!= '\n' && ch
!= '$')
1300 if (ch
== '$' && !is_foreign_ident(ident
->left
.buf
)) {
1301 strbuf_setlen(&ident
->left
, sizeof(head
) - 1);
1302 strbuf_addstr(&ident
->left
, ident
->ident
);
1304 ident
->state
= IDENT_DRAINING
;
1308 if (ident
->state
< sizeof(head
) &&
1309 head
[ident
->state
] == ch
) {
1315 strbuf_add(&ident
->left
, head
, ident
->state
);
1316 if (ident
->state
== sizeof(head
) - 1) {
1317 if (ch
!= ':' && ch
!= '$') {
1318 strbuf_addch(&ident
->left
, ch
);
1324 strbuf_addch(&ident
->left
, ch
);
1325 ident
->state
= IDENT_SKIPPING
;
1327 strbuf_addstr(&ident
->left
, ident
->ident
);
1328 ident
->state
= IDENT_DRAINING
;
1333 strbuf_addch(&ident
->left
, ch
);
1334 ident
->state
= IDENT_DRAINING
;
1339 static void ident_free_fn(struct stream_filter
*filter
)
1341 struct ident_filter
*ident
= (struct ident_filter
*)filter
;
1342 strbuf_release(&ident
->left
);
1346 static struct stream_filter_vtbl ident_vtbl
= {
1351 static struct stream_filter
*ident_filter(const unsigned char *sha1
)
1353 struct ident_filter
*ident
= xmalloc(sizeof(*ident
));
1355 xsnprintf(ident
->ident
, sizeof(ident
->ident
),
1356 ": %s $", sha1_to_hex(sha1
));
1357 strbuf_init(&ident
->left
, 0);
1358 ident
->filter
.vtbl
= &ident_vtbl
;
1360 return (struct stream_filter
*)ident
;
1364 * Return an appropriately constructed filter for the path, or NULL if
1365 * the contents cannot be filtered without reading the whole thing
1368 * Note that you would be crazy to set CRLF, smuge/clean or ident to a
1369 * large binary blob you would want us not to slurp into the memory!
1371 struct stream_filter
*get_stream_filter(const char *path
, const unsigned char *sha1
)
1373 struct conv_attrs ca
;
1374 enum crlf_action crlf_action
;
1375 struct stream_filter
*filter
= NULL
;
1377 convert_attrs(&ca
, path
);
1379 if (ca
.drv
&& (ca
.drv
->smudge
|| ca
.drv
->clean
))
1383 filter
= ident_filter(sha1
);
1385 crlf_action
= input_crlf_action(ca
.crlf_action
, ca
.eol_attr
);
1387 if ((crlf_action
== CRLF_BINARY
) || (crlf_action
== CRLF_INPUT
) ||
1388 (crlf_action
== CRLF_GUESS
&& auto_crlf
== AUTO_CRLF_FALSE
))
1389 filter
= cascade_filter(filter
, &null_filter_singleton
);
1391 else if (output_eol(crlf_action
) == EOL_CRLF
&&
1392 !(crlf_action
== CRLF_AUTO
|| crlf_action
== CRLF_GUESS
))
1393 filter
= cascade_filter(filter
, lf_to_crlf_filter());
1398 void free_stream_filter(struct stream_filter
*filter
)
1400 filter
->vtbl
->free(filter
);
1403 int stream_filter(struct stream_filter
*filter
,
1404 const char *input
, size_t *isize_p
,
1405 char *output
, size_t *osize_p
)
1407 return filter
->vtbl
->filter(filter
, input
, isize_p
, output
, osize_p
);