3 #include "run-command.h"
7 * convert.c - convert a file when checking it out and checking it in.
9 * This should use the pathname to decide on whether it wants to do some
10 * more interesting conversions (automatic gzip/unzip, general format
11 * conversions etc etc), but by default it just does automatic CRLF<->LF
12 * translation when the "text" attribute or "auto_crlf" option is set.
25 /* NUL, CR, LF and CRLF counts */
26 unsigned nul
, cr
, lf
, crlf
;
28 /* These are just approximations! */
29 unsigned printable
, nonprintable
;
32 static void gather_stats(const char *buf
, unsigned long size
, struct text_stat
*stats
)
36 memset(stats
, 0, sizeof(*stats
));
38 for (i
= 0; i
< size
; i
++) {
39 unsigned char c
= buf
[i
];
42 if (i
+1 < size
&& buf
[i
+1] == '\n')
52 stats
->nonprintable
++;
55 /* BS, HT, ESC and FF */
56 case '\b': case '\t': case '\033': case '\014':
63 stats
->nonprintable
++;
70 /* If file ends with EOF then don't count this EOF as non-printable. */
71 if (size
>= 1 && buf
[size
-1] == '\032')
72 stats
->nonprintable
--;
76 * The same heuristics as diff.c::mmfile_is_binary()
78 static int is_binary(unsigned long size
, struct text_stat
*stats
)
83 if ((stats
->printable
>> 7) < stats
->nonprintable
)
86 * Other heuristics? Average line length might be relevant,
87 * as might LF vs CR vs CRLF counts..
89 * NOTE! It might be normal to have a low ratio of CRLF to LF
90 * (somebody starts with a LF-only file and edits it with an editor
91 * that adds CRLF only to lines that are added..). But do we
92 * want to support CR-only? Probably not.
97 static enum eol
output_eol(enum crlf_action crlf_action
)
99 switch (crlf_action
) {
112 if (auto_crlf
== AUTO_CRLF_TRUE
)
114 else if (auto_crlf
== AUTO_CRLF_INPUT
)
116 else if (core_eol
== EOL_UNSET
)
122 static void check_safe_crlf(const char *path
, enum crlf_action crlf_action
,
123 struct text_stat
*stats
, enum safe_crlf checksafe
)
128 if (output_eol(crlf_action
) == EOL_LF
) {
130 * CRLFs would not be restored by checkout:
131 * check if we'd remove CRLFs
134 if (checksafe
== SAFE_CRLF_WARN
)
135 warning("CRLF will be replaced by LF in %s.\nThe file will have its original line endings in your working directory.", path
);
136 else /* i.e. SAFE_CRLF_FAIL */
137 die("CRLF would be replaced by LF in %s.", path
);
139 } else if (output_eol(crlf_action
) == EOL_CRLF
) {
141 * CRLFs would be added by checkout:
142 * check if we have "naked" LFs
144 if (stats
->lf
!= stats
->crlf
) {
145 if (checksafe
== SAFE_CRLF_WARN
)
146 warning("LF will be replaced by CRLF in %s.\nThe file will have its original line endings in your working directory.", path
);
147 else /* i.e. SAFE_CRLF_FAIL */
148 die("LF would be replaced by CRLF in %s", path
);
153 static int has_cr_in_index(const char *path
)
157 enum object_type type
;
160 struct index_state
*istate
= &the_index
;
163 pos
= index_name_pos(istate
, path
, len
);
166 * We might be in the middle of a merge, in which
167 * case we would read stage #2 (ours).
171 (pos
< 0 && i
< istate
->cache_nr
&&
172 !strcmp(istate
->cache
[i
]->name
, path
));
174 if (ce_stage(istate
->cache
[i
]) == 2)
179 data
= read_sha1_file(istate
->cache
[pos
]->sha1
, &type
, &sz
);
180 if (!data
|| type
!= OBJ_BLOB
) {
185 has_cr
= memchr(data
, '\r', sz
) != NULL
;
190 static int crlf_to_git(const char *path
, const char *src
, size_t len
,
192 enum crlf_action crlf_action
, enum safe_crlf checksafe
)
194 struct text_stat stats
;
197 if (crlf_action
== CRLF_BINARY
||
198 (crlf_action
== CRLF_GUESS
&& auto_crlf
== AUTO_CRLF_FALSE
) || !len
)
201 gather_stats(src
, len
, &stats
);
203 if (crlf_action
== CRLF_AUTO
|| crlf_action
== CRLF_GUESS
) {
205 * We're currently not going to even try to convert stuff
206 * that has bare CR characters. Does anybody do that crazy
209 if (stats
.cr
!= stats
.crlf
)
213 * And add some heuristics for binary vs text, of course...
215 if (is_binary(len
, &stats
))
218 if (crlf_action
== CRLF_GUESS
) {
220 * If the file in the index has any CR in it, do not convert.
221 * This is the new safer autocrlf handling.
223 if (has_cr_in_index(path
))
228 check_safe_crlf(path
, crlf_action
, &stats
, checksafe
);
230 /* Optimization: No CR? Nothing to convert, regardless. */
234 /* only grow if not in place */
235 if (strbuf_avail(buf
) + buf
->len
< len
)
236 strbuf_grow(buf
, len
- buf
->len
);
238 if (crlf_action
== CRLF_AUTO
|| crlf_action
== CRLF_GUESS
) {
240 * If we guessed, we already know we rejected a file with
241 * lone CR, and we can strip a CR without looking at what
245 unsigned char c
= *src
++;
251 unsigned char c
= *src
++;
252 if (! (c
== '\r' && (1 < len
&& *src
== '\n')))
256 strbuf_setlen(buf
, dst
- buf
->buf
);
260 static int crlf_to_worktree(const char *path
, const char *src
, size_t len
,
261 struct strbuf
*buf
, enum crlf_action crlf_action
)
263 char *to_free
= NULL
;
264 struct text_stat stats
;
266 if (!len
|| output_eol(crlf_action
) != EOL_CRLF
)
269 gather_stats(src
, len
, &stats
);
271 /* No LF? Nothing to convert, regardless. */
275 /* Was it already in CRLF format? */
276 if (stats
.lf
== stats
.crlf
)
279 if (crlf_action
== CRLF_AUTO
|| crlf_action
== CRLF_GUESS
) {
280 if (crlf_action
== CRLF_GUESS
) {
281 /* If we have any CR or CRLF line endings, we do not touch it */
282 /* This is the new safer autocrlf-handling */
283 if (stats
.cr
> 0 || stats
.crlf
> 0)
287 /* If we have any bare CR characters, we're not going to touch it */
288 if (stats
.cr
!= stats
.crlf
)
291 if (is_binary(len
, &stats
))
295 /* are we "faking" in place editing ? */
297 to_free
= strbuf_detach(buf
, NULL
);
299 strbuf_grow(buf
, len
+ stats
.lf
- stats
.crlf
);
301 const char *nl
= memchr(src
, '\n', len
);
304 if (nl
> src
&& nl
[-1] == '\r') {
305 strbuf_add(buf
, src
, nl
+ 1 - src
);
307 strbuf_add(buf
, src
, nl
- src
);
308 strbuf_addstr(buf
, "\r\n");
313 strbuf_add(buf
, src
, len
);
319 struct filter_params
{
326 static int filter_buffer(int in
, int out
, void *data
)
329 * Spawn cmd and feed the buffer contents through its stdin.
331 struct child_process child_process
;
332 struct filter_params
*params
= (struct filter_params
*)data
;
333 int write_err
, status
;
334 const char *argv
[] = { NULL
, NULL
};
336 /* apply % substitution to cmd */
337 struct strbuf cmd
= STRBUF_INIT
;
338 struct strbuf path
= STRBUF_INIT
;
339 struct strbuf_expand_dict_entry dict
[] = {
344 /* quote the path to preserve spaces, etc. */
345 sq_quote_buf(&path
, params
->path
);
346 dict
[0].value
= path
.buf
;
348 /* expand all %f with the quoted path */
349 strbuf_expand(&cmd
, params
->cmd
, strbuf_expand_dict_cb
, &dict
);
350 strbuf_release(&path
);
354 memset(&child_process
, 0, sizeof(child_process
));
355 child_process
.argv
= argv
;
356 child_process
.use_shell
= 1;
357 child_process
.in
= -1;
358 child_process
.out
= out
;
360 if (start_command(&child_process
))
361 return error("cannot fork to run external filter %s", params
->cmd
);
363 write_err
= (write_in_full(child_process
.in
, params
->src
, params
->size
) < 0);
364 if (close(child_process
.in
))
367 error("cannot feed the input to external filter %s", params
->cmd
);
369 status
= finish_command(&child_process
);
371 error("external filter %s failed %d", params
->cmd
, status
);
373 strbuf_release(&cmd
);
374 return (write_err
|| status
);
377 static int apply_filter(const char *path
, const char *src
, size_t len
,
378 struct strbuf
*dst
, const char *cmd
)
381 * Create a pipeline to have the command filter the buffer's
384 * (child --> cmd) --> us
387 struct strbuf nbuf
= STRBUF_INIT
;
389 struct filter_params params
;
394 memset(&async
, 0, sizeof(async
));
395 async
.proc
= filter_buffer
;
396 async
.data
= ¶ms
;
404 if (start_async(&async
))
405 return 0; /* error was already reported */
407 if (strbuf_read(&nbuf
, async
.out
, len
) < 0) {
408 error("read from external filter %s failed", cmd
);
411 if (close(async
.out
)) {
412 error("read from external filter %s failed", cmd
);
415 if (finish_async(&async
)) {
416 error("external filter %s failed", cmd
);
421 strbuf_swap(dst
, &nbuf
);
423 strbuf_release(&nbuf
);
427 static struct convert_driver
{
429 struct convert_driver
*next
;
432 } *user_convert
, **user_convert_tail
;
434 static int read_convert_config(const char *var
, const char *value
, void *cb
)
436 const char *ep
, *name
;
438 struct convert_driver
*drv
;
441 * External conversion drivers are configured using
442 * "filter.<name>.variable".
444 if (prefixcmp(var
, "filter.") || (ep
= strrchr(var
, '.')) == var
+ 6)
448 for (drv
= user_convert
; drv
; drv
= drv
->next
)
449 if (!strncmp(drv
->name
, name
, namelen
) && !drv
->name
[namelen
])
452 drv
= xcalloc(1, sizeof(struct convert_driver
));
453 drv
->name
= xmemdupz(name
, namelen
);
454 *user_convert_tail
= drv
;
455 user_convert_tail
= &(drv
->next
);
461 * filter.<name>.smudge and filter.<name>.clean specifies
466 * The command-line will not be interpolated in any way.
469 if (!strcmp("smudge", ep
))
470 return git_config_string(&drv
->smudge
, var
, value
);
472 if (!strcmp("clean", ep
))
473 return git_config_string(&drv
->clean
, var
, value
);
478 static int count_ident(const char *cp
, unsigned long size
)
481 * "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$"
493 if (memcmp("Id", cp
, 2))
504 * "$Id: ... "; scan up to the closing dollar sign and discard.
520 static int ident_to_git(const char *path
, const char *src
, size_t len
,
521 struct strbuf
*buf
, int ident
)
525 if (!ident
|| !count_ident(src
, len
))
528 /* only grow if not in place */
529 if (strbuf_avail(buf
) + buf
->len
< len
)
530 strbuf_grow(buf
, len
- buf
->len
);
533 dollar
= memchr(src
, '$', len
);
536 memcpy(dst
, src
, dollar
+ 1 - src
);
537 dst
+= dollar
+ 1 - src
;
538 len
-= dollar
+ 1 - src
;
541 if (len
> 3 && !memcmp(src
, "Id:", 3)) {
542 dollar
= memchr(src
+ 3, '$', len
- 3);
545 if (memchr(src
+ 3, '\n', dollar
- src
- 3)) {
546 /* Line break before the next dollar. */
550 memcpy(dst
, "Id$", 3);
552 len
-= dollar
+ 1 - src
;
556 memcpy(dst
, src
, len
);
557 strbuf_setlen(buf
, dst
+ len
- buf
->buf
);
561 static int ident_to_worktree(const char *path
, const char *src
, size_t len
,
562 struct strbuf
*buf
, int ident
)
564 unsigned char sha1
[20];
565 char *to_free
= NULL
, *dollar
, *spc
;
571 cnt
= count_ident(src
, len
);
575 /* are we "faking" in place editing ? */
577 to_free
= strbuf_detach(buf
, NULL
);
578 hash_sha1_file(src
, len
, "blob", sha1
);
580 strbuf_grow(buf
, len
+ cnt
* 43);
582 /* step 1: run to the next '$' */
583 dollar
= memchr(src
, '$', len
);
586 strbuf_add(buf
, src
, dollar
+ 1 - src
);
587 len
-= dollar
+ 1 - src
;
590 /* step 2: does it looks like a bit like Id:xxx$ or Id$ ? */
591 if (len
< 3 || memcmp("Id", src
, 2))
594 /* step 3: skip over Id$ or Id:xxxxx$ */
598 } else if (src
[2] == ':') {
600 * It's possible that an expanded Id has crept its way into the
601 * repository, we cope with that by stripping the expansion out.
602 * This is probably not a good idea, since it will cause changes
603 * on checkout, which won't go away by stash, but let's keep it
606 dollar
= memchr(src
+ 3, '$', len
- 3);
608 /* incomplete keyword, no more '$', so just quit the loop */
612 if (memchr(src
+ 3, '\n', dollar
- src
- 3)) {
613 /* Line break before the next dollar. */
617 spc
= memchr(src
+ 4, ' ', dollar
- src
- 4);
618 if (spc
&& spc
< dollar
-1) {
619 /* There are spaces in unexpected places.
620 * This is probably an id from some other
621 * versioning system. Keep it for now.
626 len
-= dollar
+ 1 - src
;
629 /* it wasn't a "Id$" or "Id:xxxx$" */
633 /* step 4: substitute */
634 strbuf_addstr(buf
, "Id: ");
635 strbuf_add(buf
, sha1_to_hex(sha1
), 40);
636 strbuf_addstr(buf
, " $");
638 strbuf_add(buf
, src
, len
);
644 static int git_path_check_crlf(const char *path
, struct git_attr_check
*check
)
646 const char *value
= check
->value
;
648 if (ATTR_TRUE(value
))
650 else if (ATTR_FALSE(value
))
652 else if (ATTR_UNSET(value
))
654 else if (!strcmp(value
, "input"))
656 else if (!strcmp(value
, "auto"))
661 static int git_path_check_eol(const char *path
, struct git_attr_check
*check
)
663 const char *value
= check
->value
;
665 if (ATTR_UNSET(value
))
667 else if (!strcmp(value
, "lf"))
669 else if (!strcmp(value
, "crlf"))
674 static struct convert_driver
*git_path_check_convert(const char *path
,
675 struct git_attr_check
*check
)
677 const char *value
= check
->value
;
678 struct convert_driver
*drv
;
680 if (ATTR_TRUE(value
) || ATTR_FALSE(value
) || ATTR_UNSET(value
))
682 for (drv
= user_convert
; drv
; drv
= drv
->next
)
683 if (!strcmp(value
, drv
->name
))
688 static int git_path_check_ident(const char *path
, struct git_attr_check
*check
)
690 const char *value
= check
->value
;
692 return !!ATTR_TRUE(value
);
695 static enum crlf_action
input_crlf_action(enum crlf_action text_attr
, enum eol eol_attr
)
697 if (text_attr
== CRLF_BINARY
)
699 if (eol_attr
== EOL_LF
)
701 if (eol_attr
== EOL_CRLF
)
707 struct convert_driver
*drv
;
708 enum crlf_action crlf_action
;
713 static const char *conv_attr_name
[] = {
714 "crlf", "ident", "filter", "eol", "text",
716 #define NUM_CONV_ATTRS ARRAY_SIZE(conv_attr_name)
718 static void convert_attrs(struct conv_attrs
*ca
, const char *path
)
721 static struct git_attr_check ccheck
[NUM_CONV_ATTRS
];
723 if (!ccheck
[0].attr
) {
724 for (i
= 0; i
< NUM_CONV_ATTRS
; i
++)
725 ccheck
[i
].attr
= git_attr(conv_attr_name
[i
]);
726 user_convert_tail
= &user_convert
;
727 git_config(read_convert_config
, NULL
);
730 if (!git_checkattr(path
, NUM_CONV_ATTRS
, ccheck
)) {
731 ca
->crlf_action
= git_path_check_crlf(path
, ccheck
+ 4);
732 if (ca
->crlf_action
== CRLF_GUESS
)
733 ca
->crlf_action
= git_path_check_crlf(path
, ccheck
+ 0);
734 ca
->ident
= git_path_check_ident(path
, ccheck
+ 1);
735 ca
->drv
= git_path_check_convert(path
, ccheck
+ 2);
736 ca
->eol_attr
= git_path_check_eol(path
, ccheck
+ 3);
739 ca
->crlf_action
= CRLF_GUESS
;
740 ca
->eol_attr
= EOL_UNSET
;
745 int convert_to_git(const char *path
, const char *src
, size_t len
,
746 struct strbuf
*dst
, enum safe_crlf checksafe
)
749 const char *filter
= NULL
;
750 struct conv_attrs ca
;
752 convert_attrs(&ca
, path
);
754 filter
= ca
.drv
->clean
;
756 ret
|= apply_filter(path
, src
, len
, dst
, filter
);
761 ca
.crlf_action
= input_crlf_action(ca
.crlf_action
, ca
.eol_attr
);
762 ret
|= crlf_to_git(path
, src
, len
, dst
, ca
.crlf_action
, checksafe
);
767 return ret
| ident_to_git(path
, src
, len
, dst
, ca
.ident
);
770 static int convert_to_working_tree_internal(const char *path
, const char *src
,
771 size_t len
, struct strbuf
*dst
,
775 const char *filter
= NULL
;
776 struct conv_attrs ca
;
778 convert_attrs(&ca
, path
);
780 filter
= ca
.drv
->smudge
;
782 ret
|= ident_to_worktree(path
, src
, len
, dst
, ca
.ident
);
788 * CRLF conversion can be skipped if normalizing, unless there
789 * is a smudge filter. The filter might expect CRLFs.
791 if (filter
|| !normalizing
) {
792 ca
.crlf_action
= input_crlf_action(ca
.crlf_action
, ca
.eol_attr
);
793 ret
|= crlf_to_worktree(path
, src
, len
, dst
, ca
.crlf_action
);
799 return ret
| apply_filter(path
, src
, len
, dst
, filter
);
802 int convert_to_working_tree(const char *path
, const char *src
, size_t len
, struct strbuf
*dst
)
804 return convert_to_working_tree_internal(path
, src
, len
, dst
, 0);
807 int renormalize_buffer(const char *path
, const char *src
, size_t len
, struct strbuf
*dst
)
809 int ret
= convert_to_working_tree_internal(path
, src
, len
, dst
, 1);
814 return ret
| convert_to_git(path
, src
, len
, dst
, 0);
818 * You would be crazy to set CRLF, smuge/clean or ident to
819 * a large binary blob you would want us not to slurp into
822 int can_bypass_conversion(const char *path
)
824 struct conv_attrs ca
;
825 enum crlf_action crlf_action
;
827 convert_attrs(&ca
, path
);
830 (ca
.drv
&& (ca
.drv
->smudge
|| ca
.drv
->clean
)))
833 crlf_action
= input_crlf_action(ca
.crlf_action
, ca
.eol_attr
);
834 if ((crlf_action
== CRLF_BINARY
) || (crlf_action
== CRLF_INPUT
) ||
835 (crlf_action
== CRLF_GUESS
&& auto_crlf
== AUTO_CRLF_FALSE
))