?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
The 19th batch
[git.git] / convert.c
blobc9a31eb4f03d6caa7bedba4b5237c42344f304c5
1 #define USE_THE_REPOSITORY_VARIABLE
2
3 #include "git-compat-util.h"
4 #include "advice.h"
5 #include "config.h"
6 #include "convert.h"
7 #include "copy.h"
8 #include "gettext.h"
9 #include "hex.h"
10 #include "object-store-ll.h"
11 #include "attr.h"
12 #include "run-command.h"
13 #include "quote.h"
14 #include "read-cache-ll.h"
15 #include "sigchain.h"
16 #include "pkt-line.h"
17 #include "sub-process.h"
18 #include "trace.h"
19 #include "utf8.h"
20 #include "merge-ll.h"
21
22 /*
23 * convert.c - convert a file when checking it out and checking it in.
24 *
25 * This should use the pathname to decide on whether it wants to do some
26 * more interesting conversions (automatic gzip/unzip, general format
27 * conversions etc etc), but by default it just does automatic CRLF<->LF
28 * translation when the "text" attribute or "auto_crlf" option is set.
29 */
30
31 /* Stat bits: When BIN is set, the txt bits are unset */
32 #define CONVERT_STAT_BITS_TXT_LF 0x1
33 #define CONVERT_STAT_BITS_TXT_CRLF 0x2
34 #define CONVERT_STAT_BITS_BIN 0x4
35
36 struct text_stat {
37 /* NUL, CR, LF and CRLF counts */
38 unsigned nul, lonecr, lonelf, crlf;
39
40 /* These are just approximations! */
41 unsigned printable, nonprintable;
42 };
43
44 static void gather_stats(const char *buf, unsigned long size, struct text_stat *stats)
45 {
46 unsigned long i;
47
48 memset(stats, 0, sizeof(*stats));
49
50 for (i = 0; i < size; i++) {
51 unsigned char c = buf[i];
52 if (c == '\r') {
53 if (i+1 < size && buf[i+1] == '\n') {
54 stats->crlf++;
55 i++;
56 } else
57 stats->lonecr++;
58 continue;
59 }
60 if (c == '\n') {
61 stats->lonelf++;
62 continue;
63 }
64 if (c == 127)
65 /* DEL */
66 stats->nonprintable++;
67 else if (c < 32) {
68 switch (c) {
69 /* BS, HT, ESC and FF */
70 case '\b': case '\t': case '\033': case '\014':
71 stats->printable++;
72 break;
73 case 0:
74 stats->nul++;
75 /* fall through */
76 default:
77 stats->nonprintable++;
78 }
79 }
80 else
81 stats->printable++;
82 }
83
84 /* If file ends with EOF then don't count this EOF as non-printable. */
85 if (size >= 1 && buf[size-1] == '\032')
86 stats->nonprintable--;
87 }
88
89 /*
90 * The same heuristics as diff.c::mmfile_is_binary()
91 * We treat files with bare CR as binary
92 */
93 static int convert_is_binary(const struct text_stat *stats)
94 {
95 if (stats->lonecr)
96 return 1;
97 if (stats->nul)
98 return 1;
99 if ((stats->printable >> 7) < stats->nonprintable)
100 return 1;
101 return 0;
102 }
103
104 static unsigned int gather_convert_stats(const char *data, unsigned long size)
105 {
106 struct text_stat stats;
107 int ret = 0;
108 if (!data || !size)
109 return 0;
110 gather_stats(data, size, &stats);
111 if (convert_is_binary(&stats))
112 ret |= CONVERT_STAT_BITS_BIN;
113 if (stats.crlf)
114 ret |= CONVERT_STAT_BITS_TXT_CRLF;
115 if (stats.lonelf)
116 ret |= CONVERT_STAT_BITS_TXT_LF;
117
118 return ret;
119 }
120
121 static const char *gather_convert_stats_ascii(const char *data, unsigned long size)
122 {
123 unsigned int convert_stats = gather_convert_stats(data, size);
124
125 if (convert_stats & CONVERT_STAT_BITS_BIN)
126 return "-text";
127 switch (convert_stats) {
128 case CONVERT_STAT_BITS_TXT_LF:
129 return "lf";
130 case CONVERT_STAT_BITS_TXT_CRLF:
131 return "crlf";
132 case CONVERT_STAT_BITS_TXT_LF | CONVERT_STAT_BITS_TXT_CRLF:
133 return "mixed";
134 default:
135 return "none";
136 }
137 }
138
139 const char *get_cached_convert_stats_ascii(struct index_state *istate,
140 const char *path)
141 {
142 const char *ret;
143 unsigned long sz;
144 void *data = read_blob_data_from_index(istate, path, &sz);
145 ret = gather_convert_stats_ascii(data, sz);
146 free(data);
147 return ret;
148 }
149
150 const char *get_wt_convert_stats_ascii(const char *path)
151 {
152 const char *ret = "";
153 struct strbuf sb = STRBUF_INIT;
154 if (strbuf_read_file(&sb, path, 0) >= 0)
155 ret = gather_convert_stats_ascii(sb.buf, sb.len);
156 strbuf_release(&sb);
157 return ret;
158 }
159
160 static int text_eol_is_crlf(void)
161 {
162 if (auto_crlf == AUTO_CRLF_TRUE)
163 return 1;
164 else if (auto_crlf == AUTO_CRLF_INPUT)
165 return 0;
166 if (core_eol == EOL_CRLF)
167 return 1;
168 if (core_eol == EOL_UNSET && EOL_NATIVE == EOL_CRLF)
169 return 1;
170 return 0;
171 }
172
173 static enum eol output_eol(enum convert_crlf_action crlf_action)
174 {
175 switch (crlf_action) {
176 case CRLF_BINARY:
177 return EOL_UNSET;
178 case CRLF_TEXT_CRLF:
179 return EOL_CRLF;
180 case CRLF_TEXT_INPUT:
181 return EOL_LF;
182 case CRLF_UNDEFINED:
183 case CRLF_AUTO_CRLF:
184 return EOL_CRLF;
185 case CRLF_AUTO_INPUT:
186 return EOL_LF;
187 case CRLF_TEXT:
188 case CRLF_AUTO:
189 /* fall through */
190 return text_eol_is_crlf() ? EOL_CRLF : EOL_LF;
191 }
192 warning(_("illegal crlf_action %d"), (int)crlf_action);
193 return core_eol;
194 }
195
196 static void check_global_conv_flags_eol(const char *path,
197 struct text_stat *old_stats, struct text_stat *new_stats,
198 int conv_flags)
199 {
200 if (old_stats->crlf && !new_stats->crlf ) {
201 /*
202 * CRLFs would not be restored by checkout
203 */
204 if (conv_flags & CONV_EOL_RNDTRP_DIE)
205 die(_("CRLF would be replaced by LF in %s"), path);
206 else if (conv_flags & CONV_EOL_RNDTRP_WARN)
207 warning(_("in the working copy of '%s', CRLF will be"
208 " replaced by LF the next time Git touches"
209 " it"), path);
210 } else if (old_stats->lonelf && !new_stats->lonelf ) {
211 /*
212 * CRLFs would be added by checkout
213 */
214 if (conv_flags & CONV_EOL_RNDTRP_DIE)
215 die(_("LF would be replaced by CRLF in %s"), path);
216 else if (conv_flags & CONV_EOL_RNDTRP_WARN)
217 warning(_("in the working copy of '%s', LF will be"
218 " replaced by CRLF the next time Git touches"
219 " it"), path);
220 }
221 }
222
223 static int has_crlf_in_index(struct index_state *istate, const char *path)
224 {
225 unsigned long sz;
226 void *data;
227 const char *crp;
228 int has_crlf = 0;
229
230 data = read_blob_data_from_index(istate, path, &sz);
231 if (!data)
232 return 0;
233
234 crp = memchr(data, '\r', sz);
235 if (crp) {
236 unsigned int ret_stats;
237 ret_stats = gather_convert_stats(data, sz);
238 if (!(ret_stats & CONVERT_STAT_BITS_BIN) &&
239 (ret_stats & CONVERT_STAT_BITS_TXT_CRLF))
240 has_crlf = 1;
241 }
242 free(data);
243 return has_crlf;
244 }
245
246 static int will_convert_lf_to_crlf(struct text_stat *stats,
247 enum convert_crlf_action crlf_action)
248 {
249 if (output_eol(crlf_action) != EOL_CRLF)
250 return 0;
251 /* No "naked" LF? Nothing to convert, regardless. */
252 if (!stats->lonelf)
253 return 0;
254
255 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
256 /* If we have any CR or CRLF line endings, we do not touch it */
257 /* This is the new safer autocrlf-handling */
258 if (stats->lonecr || stats->crlf)
259 return 0;
260
261 if (convert_is_binary(stats))
262 return 0;
263 }
264 return 1;
265
266 }
267
268 static int validate_encoding(const char *path, const char *enc,
269 const char *data, size_t len, int die_on_error)
270 {
271 const char *stripped;
272
273 /* We only check for UTF here as UTF?? can be an alias for UTF-?? */
274 if (skip_iprefix(enc, "UTF", &stripped)) {
275 skip_prefix(stripped, "-", &stripped);
276
277 /*
278 * Check for detectable errors in UTF encodings
279 */
280 if (has_prohibited_utf_bom(enc, data, len)) {
281 const char *error_msg = _(
282 "BOM is prohibited in '%s' if encoded as %s");
283 /*
284 * This advice is shown for UTF-??BE and UTF-??LE encodings.
285 * We cut off the last two characters of the encoding name
286 * to generate the encoding name suitable for BOMs.
287 */
288 const char *advise_msg = _(
289 "The file '%s' contains a byte order "
290 "mark (BOM). Please use UTF-%.*s as "
291 "working-tree-encoding.");
292 int stripped_len = strlen(stripped) - strlen("BE");
293 advise(advise_msg, path, stripped_len, stripped);
294 if (die_on_error)
295 die(error_msg, path, enc);
296 else {
297 return error(error_msg, path, enc);
298 }
299
300 } else if (is_missing_required_utf_bom(enc, data, len)) {
301 const char *error_msg = _(
302 "BOM is required in '%s' if encoded as %s");
303 const char *advise_msg = _(
304 "The file '%s' is missing a byte order "
305 "mark (BOM). Please use UTF-%sBE or UTF-%sLE "
306 "(depending on the byte order) as "
307 "working-tree-encoding.");
308 advise(advise_msg, path, stripped, stripped);
309 if (die_on_error)
310 die(error_msg, path, enc);
311 else {
312 return error(error_msg, path, enc);
313 }
314 }
315
316 }
317 return 0;
318 }
319
320 static void trace_encoding(const char *context, const char *path,
321 const char *encoding, const char *buf, size_t len)
322 {
323 static struct trace_key coe = TRACE_KEY_INIT(WORKING_TREE_ENCODING);
324 struct strbuf trace = STRBUF_INIT;
325 int i;
326
327 if (!trace_want(&coe))
328 return;
329
330 strbuf_addf(&trace, "%s (%s, considered %s):\n", context, path, encoding);
331 for (i = 0; i < len && buf; ++i) {
332 strbuf_addf(
333 &trace, "| \033[2m%2i:\033[0m %2x \033[2m%c\033[0m%c",
334 i,
335 (unsigned char) buf[i],
336 (buf[i] > 32 && buf[i] < 127 ? buf[i] : ' '),
337 ((i+1) % 8 && (i+1) < len ? ' ' : '\n')
338 );
339 }
340 strbuf_addchars(&trace, '\n', 1);
341
342 trace_strbuf(&coe, &trace);
343 strbuf_release(&trace);
344 }
345
346 static int check_roundtrip(const char *enc_name)
347 {
348 /*
349 * check_roundtrip_encoding contains a string of comma and/or
350 * space separated encodings (eg. "UTF-16, ASCII, CP1125").
351 * Search for the given encoding in that string.
352 */
353 const char *encoding = check_roundtrip_encoding ?
354 check_roundtrip_encoding : "SHIFT-JIS";
355 const char *found = strcasestr(encoding, enc_name);
356 const char *next;
357 int len;
358 if (!found)
359 return 0;
360 next = found + strlen(enc_name);
361 len = strlen(encoding);
362 return (found && (
363 /*
364 * Check that the found encoding is at the beginning of
365 * encoding or that it is prefixed with a space or
366 * comma.
367 */
368 found == encoding || (
369 (isspace(found[-1]) || found[-1] == ',')
370 )
371 ) && (
372 /*
373 * Check that the found encoding is at the end of
374 * encoding or that it is suffixed with a space
375 * or comma.
376 */
377 next == encoding + len || (
378 next < encoding + len &&
379 (isspace(next[0]) || next[0] == ',')
380 )
381 ));
382 }
383
384 static const char *default_encoding = "UTF-8";
385
386 static int encode_to_git(const char *path, const char *src, size_t src_len,
387 struct strbuf *buf, const char *enc, int conv_flags)
388 {
389 char *dst;
390 size_t dst_len;
391 int die_on_error = conv_flags & CONV_WRITE_OBJECT;
392
393 /*
394 * No encoding is specified or there is nothing to encode.
395 * Tell the caller that the content was not modified.
396 */
397 if (!enc || (src && !src_len))
398 return 0;
399
400 /*
401 * Looks like we got called from "would_convert_to_git()".
402 * This means Git wants to know if it would encode (= modify!)
403 * the content. Let's answer with "yes", since an encoding was
404 * specified.
405 */
406 if (!buf && !src)
407 return 1;
408
409 if (validate_encoding(path, enc, src, src_len, die_on_error))
410 return 0;
411
412 trace_encoding("source", path, enc, src, src_len);
413 dst = reencode_string_len(src, src_len, default_encoding, enc,
414 &dst_len);
415 if (!dst) {
416 /*
417 * We could add the blob "as-is" to Git. However, on checkout
418 * we would try to re-encode to the original encoding. This
419 * would fail and we would leave the user with a messed-up
420 * working tree. Let's try to avoid this by screaming loud.
421 */
422 const char* msg = _("failed to encode '%s' from %s to %s");
423 if (die_on_error)
424 die(msg, path, enc, default_encoding);
425 else {
426 error(msg, path, enc, default_encoding);
427 return 0;
428 }
429 }
430 trace_encoding("destination", path, default_encoding, dst, dst_len);
431
432 /*
433 * UTF supports lossless conversion round tripping [1] and conversions
434 * between UTF and other encodings are mostly round trip safe as
435 * Unicode aims to be a superset of all other character encodings.
436 * However, certain encodings (e.g. SHIFT-JIS) are known to have round
437 * trip issues [2]. Check the round trip conversion for all encodings
438 * listed in core.checkRoundtripEncoding.
439 *
440 * The round trip check is only performed if content is written to Git.
441 * This ensures that no information is lost during conversion to/from
442 * the internal UTF-8 representation.
443 *
444 * Please note, the code below is not tested because I was not able to
445 * generate a faulty round trip without an iconv error. Iconv errors
446 * are already caught above.
447 *
448 * [1] http://unicode.org/faq/utf_bom.html#gen2
449 * [2] https://support.microsoft.com/en-us/help/170559/prb-conversion-problem-between-shift-jis-and-unicode
450 */
451 if (die_on_error && check_roundtrip(enc)) {
452 char *re_src;
453 size_t re_src_len;
454
455 re_src = reencode_string_len(dst, dst_len,
456 enc, default_encoding,
457 &re_src_len);
458
459 trace_printf("Checking roundtrip encoding for %s...\n", enc);
460 trace_encoding("reencoded source", path, enc,
461 re_src, re_src_len);
462
463 if (!re_src || src_len != re_src_len ||
464 memcmp(src, re_src, src_len)) {
465 const char* msg = _("encoding '%s' from %s to %s and "
466 "back is not the same");
467 die(msg, path, enc, default_encoding);
468 }
469
470 free(re_src);
471 }
472
473 strbuf_attach(buf, dst, dst_len, dst_len + 1);
474 return 1;
475 }
476
477 static int encode_to_worktree(const char *path, const char *src, size_t src_len,
478 struct strbuf *buf, const char *enc)
479 {
480 char *dst;
481 size_t dst_len;
482
483 /*
484 * No encoding is specified or there is nothing to encode.
485 * Tell the caller that the content was not modified.
486 */
487 if (!enc || (src && !src_len))
488 return 0;
489
490 dst = reencode_string_len(src, src_len, enc, default_encoding,
491 &dst_len);
492 if (!dst) {
493 error(_("failed to encode '%s' from %s to %s"),
494 path, default_encoding, enc);
495 return 0;
496 }
497
498 strbuf_attach(buf, dst, dst_len, dst_len + 1);
499 return 1;
500 }
501
502 static int crlf_to_git(struct index_state *istate,
503 const char *path, const char *src, size_t len,
504 struct strbuf *buf,
505 enum convert_crlf_action crlf_action, int conv_flags)
506 {
507 struct text_stat stats;
508 char *dst;
509 int convert_crlf_into_lf;
510
511 if (crlf_action == CRLF_BINARY ||
512 (src && !len))
513 return 0;
514
515 /*
516 * If we are doing a dry-run and have no source buffer, there is
517 * nothing to analyze; we must assume we would convert.
518 */
519 if (!buf && !src)
520 return 1;
521
522 gather_stats(src, len, &stats);
523 /* Optimization: No CRLF? Nothing to convert, regardless. */
524 convert_crlf_into_lf = !!stats.crlf;
525
526 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
527 if (convert_is_binary(&stats))
528 return 0;
529 /*
530 * If the file in the index has any CR in it, do not
531 * convert. This is the new safer autocrlf handling,
532 * unless we want to renormalize in a merge or
533 * cherry-pick.
534 */
535 if ((!(conv_flags & CONV_EOL_RENORMALIZE)) &&
536 has_crlf_in_index(istate, path))
537 convert_crlf_into_lf = 0;
538 }
539 if (((conv_flags & CONV_EOL_RNDTRP_WARN) ||
540 ((conv_flags & CONV_EOL_RNDTRP_DIE) && len))) {
541 struct text_stat new_stats;
542 memcpy(&new_stats, &stats, sizeof(new_stats));
543 /* simulate "git add" */
544 if (convert_crlf_into_lf) {
545 new_stats.lonelf += new_stats.crlf;
546 new_stats.crlf = 0;
547 }
548 /* simulate "git checkout" */
549 if (will_convert_lf_to_crlf(&new_stats, crlf_action)) {
550 new_stats.crlf += new_stats.lonelf;
551 new_stats.lonelf = 0;
552 }
553 check_global_conv_flags_eol(path, &stats, &new_stats, conv_flags);
554 }
555 if (!convert_crlf_into_lf)
556 return 0;
557
558 /*
559 * At this point all of our source analysis is done, and we are sure we
560 * would convert. If we are in dry-run mode, we can give an answer.
561 */
562 if (!buf)
563 return 1;
564
565 /* only grow if not in place */
566 if (strbuf_avail(buf) + buf->len < len)
567 strbuf_grow(buf, len - buf->len);
568 dst = buf->buf;
569 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
570 /*
571 * If we guessed, we already know we rejected a file with
572 * lone CR, and we can strip a CR without looking at what
573 * follow it.
574 */
575 do {
576 unsigned char c = *src++;
577 if (c != '\r')
578 *dst++ = c;
579 } while (--len);
580 } else {
581 do {
582 unsigned char c = *src++;
583 if (! (c == '\r' && (1 < len && *src == '\n')))
584 *dst++ = c;
585 } while (--len);
586 }
587 strbuf_setlen(buf, dst - buf->buf);
588 return 1;
589 }
590
591 static int crlf_to_worktree(const char *src, size_t len, struct strbuf *buf,
592 enum convert_crlf_action crlf_action)
593 {
594 char *to_free = NULL;
595 struct text_stat stats;
596
597 if (!len || output_eol(crlf_action) != EOL_CRLF)
598 return 0;
599
600 gather_stats(src, len, &stats);
601 if (!will_convert_lf_to_crlf(&stats, crlf_action))
602 return 0;
603
604 /* are we "faking" in place editing ? */
605 if (src == buf->buf)
606 to_free = strbuf_detach(buf, NULL);
607
608 strbuf_grow(buf, len + stats.lonelf);
609 for (;;) {
610 const char *nl = memchr(src, '\n', len);
611 if (!nl)
612 break;
613 if (nl > src && nl[-1] == '\r') {
614 strbuf_add(buf, src, nl + 1 - src);
615 } else {
616 strbuf_add(buf, src, nl - src);
617 strbuf_addstr(buf, "\r\n");
618 }
619 len -= nl + 1 - src;
620 src = nl + 1;
621 }
622 strbuf_add(buf, src, len);
623
624 free(to_free);
625 return 1;
626 }
627
628 struct filter_params {
629 const char *src;
630 size_t size;
631 int fd;
632 const char *cmd;
633 const char *path;
634 };
635
636 static int filter_buffer_or_fd(int in UNUSED, int out, void *data)
637 {
638 /*
639 * Spawn cmd and feed the buffer contents through its stdin.
640 */
641 struct child_process child_process = CHILD_PROCESS_INIT;
642 struct filter_params *params = (struct filter_params *)data;
643 const char *format = params->cmd;
644 int write_err, status;
645
646 /* apply % substitution to cmd */
647 struct strbuf cmd = STRBUF_INIT;
648
649 /* expand all %f with the quoted path; quote to preserve space, etc. */
650 while (strbuf_expand_step(&cmd, &format)) {
651 if (skip_prefix(format, "%", &format))
652 strbuf_addch(&cmd, '%');
653 else if (skip_prefix(format, "f", &format))
654 sq_quote_buf(&cmd, params->path);
655 else
656 strbuf_addch(&cmd, '%');
657 }
658
659 strvec_push(&child_process.args, cmd.buf);
660 child_process.use_shell = 1;
661 child_process.in = -1;
662 child_process.out = out;
663
664 if (start_command(&child_process)) {
665 strbuf_release(&cmd);
666 return error(_("cannot fork to run external filter '%s'"),
667 params->cmd);
668 }
669
670 sigchain_push(SIGPIPE, SIG_IGN);
671
672 if (params->src) {
673 write_err = (write_in_full(child_process.in,
674 params->src, params->size) < 0);
675 if (errno == EPIPE)
676 write_err = 0;
677 } else {
678 write_err = copy_fd(params->fd, child_process.in);
679 if (write_err == COPY_WRITE_ERROR && errno == EPIPE)
680 write_err = 0;
681 }
682
683 if (close(child_process.in))
684 write_err = 1;
685 if (write_err)
686 error(_("cannot feed the input to external filter '%s'"),
687 params->cmd);
688
689 sigchain_pop(SIGPIPE);
690
691 status = finish_command(&child_process);
692 if (status)
693 error(_("external filter '%s' failed %d"), params->cmd, status);
694
695 strbuf_release(&cmd);
696 return (write_err || status);
697 }
698
699 static int apply_single_file_filter(const char *path, const char *src, size_t len, int fd,
700 struct strbuf *dst, const char *cmd)
701 {
702 /*
703 * Create a pipeline to have the command filter the buffer's
704 * contents.
705 *
706 * (child --> cmd) --> us
707 */
708 int err = 0;
709 struct strbuf nbuf = STRBUF_INIT;
710 struct async async;
711 struct filter_params params;
712
713 memset(&async, 0, sizeof(async));
714 async.proc = filter_buffer_or_fd;
715 async.data = &params;
716 async.out = -1;
717 params.src = src;
718 params.size = len;
719 params.fd = fd;
720 params.cmd = cmd;
721 params.path = path;
722
723 fflush(NULL);
724 if (start_async(&async))
725 return 0; /* error was already reported */
726
727 if (strbuf_read(&nbuf, async.out, 0) < 0) {
728 err = error(_("read from external filter '%s' failed"), cmd);
729 }
730 if (close(async.out)) {
731 err = error(_("read from external filter '%s' failed"), cmd);
732 }
733 if (finish_async(&async)) {
734 err = error(_("external filter '%s' failed"), cmd);
735 }
736
737 if (!err) {
738 strbuf_swap(dst, &nbuf);
739 }
740 strbuf_release(&nbuf);
741 return !err;
742 }
743
744 #define CAP_CLEAN (1u<<0)
745 #define CAP_SMUDGE (1u<<1)
746 #define CAP_DELAY (1u<<2)
747
748 struct cmd2process {
749 struct subprocess_entry subprocess; /* must be the first member! */
750 unsigned int supported_capabilities;
751 };
752
753 static int subprocess_map_initialized;
754 static struct hashmap subprocess_map;
755
756 static int start_multi_file_filter_fn(struct subprocess_entry *subprocess)
757 {
758 static int versions[] = {2, 0};
759 static struct subprocess_capability capabilities[] = {
760 { "clean", CAP_CLEAN },
761 { "smudge", CAP_SMUDGE },
762 { "delay", CAP_DELAY },
763 { NULL, 0 }
764 };
765 struct cmd2process *entry = (struct cmd2process *)subprocess;
766 return subprocess_handshake(subprocess, "git-filter", versions, NULL,
767 capabilities,
768 &entry->supported_capabilities);
769 }
770
771 static void handle_filter_error(const struct strbuf *filter_status,
772 struct cmd2process *entry,
773 const unsigned int wanted_capability)
774 {
775 if (!strcmp(filter_status->buf, "error"))
776 ; /* The filter signaled a problem with the file. */
777 else if (!strcmp(filter_status->buf, "abort") && wanted_capability) {
778 /*
779 * The filter signaled a permanent problem. Don't try to filter
780 * files with the same command for the lifetime of the current
781 * Git process.
782 */
783 entry->supported_capabilities &= ~wanted_capability;
784 } else {
785 /*
786 * Something went wrong with the protocol filter.
787 * Force shutdown and restart if another blob requires filtering.
788 */
789 error(_("external filter '%s' failed"), entry->subprocess.cmd);
790 subprocess_stop(&subprocess_map, &entry->subprocess);
791 free(entry);
792 }
793 }
794
795 static int apply_multi_file_filter(const char *path, const char *src, size_t len,
796 int fd, struct strbuf *dst, const char *cmd,
797 const unsigned int wanted_capability,
798 const struct checkout_metadata *meta,
799 struct delayed_checkout *dco)
800 {
801 int err;
802 int can_delay = 0;
803 struct cmd2process *entry;
804 struct child_process *process;
805 struct strbuf nbuf = STRBUF_INIT;
806 struct strbuf filter_status = STRBUF_INIT;
807 const char *filter_type;
808
809 if (!subprocess_map_initialized) {
810 subprocess_map_initialized = 1;
811 hashmap_init(&subprocess_map, cmd2process_cmp, NULL, 0);
812 entry = NULL;
813 } else {
814 entry = (struct cmd2process *)subprocess_find_entry(&subprocess_map, cmd);
815 }
816
817 fflush(NULL);
818
819 if (!entry) {
820 entry = xmalloc(sizeof(*entry));
821 entry->supported_capabilities = 0;
822
823 if (subprocess_start(&subprocess_map, &entry->subprocess, cmd, start_multi_file_filter_fn)) {
824 free(entry);
825 return 0;
826 }
827 }
828 process = &entry->subprocess.process;
829
830 if (!(entry->supported_capabilities & wanted_capability))
831 return 0;
832
833 if (wanted_capability & CAP_CLEAN)
834 filter_type = "clean";
835 else if (wanted_capability & CAP_SMUDGE)
836 filter_type = "smudge";
837 else
838 die(_("unexpected filter type"));
839
840 sigchain_push(SIGPIPE, SIG_IGN);
841
842 assert(strlen(filter_type) < LARGE_PACKET_DATA_MAX - strlen("command=\n"));
843 err = packet_write_fmt_gently(process->in, "command=%s\n", filter_type);
844 if (err)
845 goto done;
846
847 err = strlen(path) > LARGE_PACKET_DATA_MAX - strlen("pathname=\n");
848 if (err) {
849 error(_("path name too long for external filter"));
850 goto done;
851 }
852
853 err = packet_write_fmt_gently(process->in, "pathname=%s\n", path);
854 if (err)
855 goto done;
856
857 if (meta && meta->refname) {
858 err = packet_write_fmt_gently(process->in, "ref=%s\n", meta->refname);
859 if (err)
860 goto done;
861 }
862
863 if (meta && !is_null_oid(&meta->treeish)) {
864 err = packet_write_fmt_gently(process->in, "treeish=%s\n", oid_to_hex(&meta->treeish));
865 if (err)
866 goto done;
867 }
868
869 if (meta && !is_null_oid(&meta->blob)) {
870 err = packet_write_fmt_gently(process->in, "blob=%s\n", oid_to_hex(&meta->blob));
871 if (err)
872 goto done;
873 }
874
875 if ((entry->supported_capabilities & CAP_DELAY) &&
876 dco && dco->state == CE_CAN_DELAY) {
877 can_delay = 1;
878 err = packet_write_fmt_gently(process->in, "can-delay=1\n");
879 if (err)
880 goto done;
881 }
882
883 err = packet_flush_gently(process->in);
884 if (err)
885 goto done;
886
887 if (fd >= 0)
888 err = write_packetized_from_fd_no_flush(fd, process->in);
889 else
890 err = write_packetized_from_buf_no_flush(src, len, process->in);
891 if (err)
892 goto done;
893
894 err = packet_flush_gently(process->in);
895 if (err)
896 goto done;
897
898 err = subprocess_read_status(process->out, &filter_status);
899 if (err)
900 goto done;
901
902 if (can_delay && !strcmp(filter_status.buf, "delayed")) {
903 string_list_insert(&dco->filters, cmd);
904 string_list_insert(&dco->paths, path);
905 } else {
906 /* The filter got the blob and wants to send us a response. */
907 err = strcmp(filter_status.buf, "success");
908 if (err)
909 goto done;
910
911 err = read_packetized_to_strbuf(process->out, &nbuf,
912 PACKET_READ_GENTLE_ON_EOF) < 0;
913 if (err)
914 goto done;
915
916 err = subprocess_read_status(process->out, &filter_status);
917 if (err)
918 goto done;
919
920 err = strcmp(filter_status.buf, "success");
921 }
922
923 done:
924 sigchain_pop(SIGPIPE);
925
926 if (err)
927 handle_filter_error(&filter_status, entry, wanted_capability);
928 else
929 strbuf_swap(dst, &nbuf);
930 strbuf_release(&nbuf);
931 strbuf_release(&filter_status);
932 return !err;
933 }
934
935
936 int async_query_available_blobs(const char *cmd, struct string_list *available_paths)
937 {
938 int err;
939 char *line;
940 struct cmd2process *entry;
941 struct child_process *process;
942 struct strbuf filter_status = STRBUF_INIT;
943
944 assert(subprocess_map_initialized);
945 entry = (struct cmd2process *)subprocess_find_entry(&subprocess_map, cmd);
946 if (!entry) {
947 error(_("external filter '%s' is not available anymore although "
948 "not all paths have been filtered"), cmd);
949 return 0;
950 }
951 process = &entry->subprocess.process;
952 sigchain_push(SIGPIPE, SIG_IGN);
953
954 err = packet_write_fmt_gently(
955 process->in, "command=list_available_blobs\n");
956 if (err)
957 goto done;
958
959 err = packet_flush_gently(process->in);
960 if (err)
961 goto done;
962
963 while ((line = packet_read_line(process->out, NULL))) {
964 const char *path;
965 if (skip_prefix(line, "pathname=", &path))
966 string_list_insert(available_paths, path);
967 else
968 ; /* ignore unknown keys */
969 }
970
971 err = subprocess_read_status(process->out, &filter_status);
972 if (err)
973 goto done;
974
975 err = strcmp(filter_status.buf, "success");
976
977 done:
978 sigchain_pop(SIGPIPE);
979
980 if (err)
981 handle_filter_error(&filter_status, entry, 0);
982 strbuf_release(&filter_status);
983 return !err;
984 }
985
986 static struct convert_driver {
987 const char *name;
988 struct convert_driver *next;
989 char *smudge;
990 char *clean;
991 char *process;
992 int required;
993 } *user_convert, **user_convert_tail;
994
995 static int apply_filter(const char *path, const char *src, size_t len,
996 int fd, struct strbuf *dst, struct convert_driver *drv,
997 const unsigned int wanted_capability,
998 const struct checkout_metadata *meta,
999 struct delayed_checkout *dco)
1000 {
1001 const char *cmd = NULL;
1002
1003 if (!drv)
1004 return 0;
1005
1006 if (!dst)
1007 return 1;
1008
1009 if ((wanted_capability & CAP_CLEAN) && !drv->process && drv->clean)
1010 cmd = drv->clean;
1011 else if ((wanted_capability & CAP_SMUDGE) && !drv->process && drv->smudge)
1012 cmd = drv->smudge;
1013
1014 if (cmd && *cmd)
1015 return apply_single_file_filter(path, src, len, fd, dst, cmd);
1016 else if (drv->process && *drv->process)
1017 return apply_multi_file_filter(path, src, len, fd, dst,
1018 drv->process, wanted_capability, meta, dco);
1019
1020 return 0;
1021 }
1022
1023 static int read_convert_config(const char *var, const char *value,
1024 const struct config_context *ctx UNUSED,
1025 void *cb UNUSED)
1026 {
1027 const char *key, *name;
1028 size_t namelen;
1029 struct convert_driver *drv;
1030
1031 /*
1032 * External conversion drivers are configured using
1033 * "filter.<name>.variable".
1034 */
1035 if (parse_config_key(var, "filter", &name, &namelen, &key) < 0 || !name)
1036 return 0;
1037 for (drv = user_convert; drv; drv = drv->next)
1038 if (!xstrncmpz(drv->name, name, namelen))
1039 break;
1040 if (!drv) {
1041 CALLOC_ARRAY(drv, 1);
1042 drv->name = xmemdupz(name, namelen);
1043 *user_convert_tail = drv;
1044 user_convert_tail = &(drv->next);
1045 }
1046
1047 /*
1048 * filter.<name>.smudge and filter.<name>.clean specifies
1049 * the command line:
1050 *
1051 * command-line
1052 *
1053 * The command-line will not be interpolated in any way.
1054 */
1055
1056 if (!strcmp("smudge", key)) {
1057 FREE_AND_NULL(drv->smudge);
1058 return git_config_string(&drv->smudge, var, value);
1059 }
1060
1061 if (!strcmp("clean", key)) {
1062 FREE_AND_NULL(drv->clean);
1063 return git_config_string(&drv->clean, var, value);
1064 }
1065
1066 if (!strcmp("process", key)) {
1067 FREE_AND_NULL(drv->process);
1068 return git_config_string(&drv->process, var, value);
1069 }
1070
1071 if (!strcmp("required", key)) {
1072 drv->required = git_config_bool(var, value);
1073 return 0;
1074 }
1075
1076 return 0;
1077 }
1078
1079 static int count_ident(const char *cp, unsigned long size)
1080 {
1081 /*
1082 * "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$"
1083 */
1084 int cnt = 0;
1085 char ch;
1086
1087 while (size) {
1088 ch = *cp++;
1089 size--;
1090 if (ch != '$')
1091 continue;
1092 if (size < 3)
1093 break;
1094 if (memcmp("Id", cp, 2))
1095 continue;
1096 ch = cp[2];
1097 cp += 3;
1098 size -= 3;
1099 if (ch == '$')
1100 cnt++; /* $Id$ */
1101 if (ch != ':')
1102 continue;
1103
1104 /*
1105 * "$Id: ... "; scan up to the closing dollar sign and discard.
1106 */
1107 while (size) {
1108 ch = *cp++;
1109 size--;
1110 if (ch == '$') {
1111 cnt++;
1112 break;
1113 }
1114 if (ch == '\n')
1115 break;
1116 }
1117 }
1118 return cnt;
1119 }
1120
1121 static int ident_to_git(const char *src, size_t len,
1122 struct strbuf *buf, int ident)
1123 {
1124 char *dst, *dollar;
1125
1126 if (!ident || (src && !count_ident(src, len)))
1127 return 0;
1128
1129 if (!buf)
1130 return 1;
1131
1132 /* only grow if not in place */
1133 if (strbuf_avail(buf) + buf->len < len)
1134 strbuf_grow(buf, len - buf->len);
1135 dst = buf->buf;
1136 for (;;) {
1137 dollar = memchr(src, '$', len);
1138 if (!dollar)
1139 break;
1140 memmove(dst, src, dollar + 1 - src);
1141 dst += dollar + 1 - src;
1142 len -= dollar + 1 - src;
1143 src = dollar + 1;
1144
1145 if (len > 3 && !memcmp(src, "Id:", 3)) {
1146 dollar = memchr(src + 3, '$', len - 3);
1147 if (!dollar)
1148 break;
1149 if (memchr(src + 3, '\n', dollar - src - 3)) {
1150 /* Line break before the next dollar. */
1151 continue;
1152 }
1153
1154 memcpy(dst, "Id$", 3);
1155 dst += 3;
1156 len -= dollar + 1 - src;
1157 src = dollar + 1;
1158 }
1159 }
1160 memmove(dst, src, len);
1161 strbuf_setlen(buf, dst + len - buf->buf);
1162 return 1;
1163 }
1164
1165 static int ident_to_worktree(const char *src, size_t len,
1166 struct strbuf *buf, int ident)
1167 {
1168 struct object_id oid;
1169 char *to_free = NULL, *dollar, *spc;
1170 int cnt;
1171
1172 if (!ident)
1173 return 0;
1174
1175 cnt = count_ident(src, len);
1176 if (!cnt)
1177 return 0;
1178
1179 /* are we "faking" in place editing ? */
1180 if (src == buf->buf)
1181 to_free = strbuf_detach(buf, NULL);
1182 hash_object_file(the_hash_algo, src, len, OBJ_BLOB, &oid);
1183
1184 strbuf_grow(buf, len + cnt * (the_hash_algo->hexsz + 3));
1185 for (;;) {
1186 /* step 1: run to the next '$' */
1187 dollar = memchr(src, '$', len);
1188 if (!dollar)
1189 break;
1190 strbuf_add(buf, src, dollar + 1 - src);
1191 len -= dollar + 1 - src;
1192 src = dollar + 1;
1193
1194 /* step 2: does it looks like a bit like Id:xxx$ or Id$ ? */
1195 if (len < 3 || memcmp("Id", src, 2))
1196 continue;
1197
1198 /* step 3: skip over Id$ or Id:xxxxx$ */
1199 if (src[2] == '$') {
1200 src += 3;
1201 len -= 3;
1202 } else if (src[2] == ':') {
1203 /*
1204 * It's possible that an expanded Id has crept its way into the
1205 * repository, we cope with that by stripping the expansion out.
1206 * This is probably not a good idea, since it will cause changes
1207 * on checkout, which won't go away by stash, but let's keep it
1208 * for git-style ids.
1209 */
1210 dollar = memchr(src + 3, '$', len - 3);
1211 if (!dollar) {
1212 /* incomplete keyword, no more '$', so just quit the loop */
1213 break;
1214 }
1215
1216 if (memchr(src + 3, '\n', dollar - src - 3)) {
1217 /* Line break before the next dollar. */
1218 continue;
1219 }
1220
1221 spc = memchr(src + 4, ' ', dollar - src - 4);
1222 if (spc && spc < dollar-1) {
1223 /* There are spaces in unexpected places.
1224 * This is probably an id from some other
1225 * versioning system. Keep it for now.
1226 */
1227 continue;
1228 }
1229
1230 len -= dollar + 1 - src;
1231 src = dollar + 1;
1232 } else {
1233 /* it wasn't a "Id$" or "Id:xxxx$" */
1234 continue;
1235 }
1236
1237 /* step 4: substitute */
1238 strbuf_addstr(buf, "Id: ");
1239 strbuf_addstr(buf, oid_to_hex(&oid));
1240 strbuf_addstr(buf, " $");
1241 }
1242 strbuf_add(buf, src, len);
1243
1244 free(to_free);
1245 return 1;
1246 }
1247
1248 static const char *git_path_check_encoding(struct attr_check_item *check)
1249 {
1250 const char *value = check->value;
1251
1252 if (ATTR_UNSET(value) || !strlen(value))
1253 return NULL;
1254
1255 if (ATTR_TRUE(value) || ATTR_FALSE(value)) {
1256 die(_("true/false are no valid working-tree-encodings"));
1257 }
1258
1259 /* Don't encode to the default encoding */
1260 if (same_encoding(value, default_encoding))
1261 return NULL;
1262
1263 return value;
1264 }
1265
1266 static enum convert_crlf_action git_path_check_crlf(struct attr_check_item *check)
1267 {
1268 const char *value = check->value;
1269
1270 if (ATTR_TRUE(value))
1271 return CRLF_TEXT;
1272 else if (ATTR_FALSE(value))
1273 return CRLF_BINARY;
1274 else if (ATTR_UNSET(value))
1275 ;
1276 else if (!strcmp(value, "input"))
1277 return CRLF_TEXT_INPUT;
1278 else if (!strcmp(value, "auto"))
1279 return CRLF_AUTO;
1280 return CRLF_UNDEFINED;
1281 }
1282
1283 static enum eol git_path_check_eol(struct attr_check_item *check)
1284 {
1285 const char *value = check->value;
1286
1287 if (ATTR_UNSET(value))
1288 ;
1289 else if (!strcmp(value, "lf"))
1290 return EOL_LF;
1291 else if (!strcmp(value, "crlf"))
1292 return EOL_CRLF;
1293 return EOL_UNSET;
1294 }
1295
1296 static struct convert_driver *git_path_check_convert(struct attr_check_item *check)
1297 {
1298 const char *value = check->value;
1299 struct convert_driver *drv;
1300
1301 if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
1302 return NULL;
1303 for (drv = user_convert; drv; drv = drv->next)
1304 if (!strcmp(value, drv->name))
1305 return drv;
1306 return NULL;
1307 }
1308
1309 static int git_path_check_ident(struct attr_check_item *check)
1310 {
1311 const char *value = check->value;
1312
1313 return !!ATTR_TRUE(value);
1314 }
1315
1316 static struct attr_check *check;
1317
1318 void convert_attrs(struct index_state *istate,
1319 struct conv_attrs *ca, const char *path)
1320 {
1321 struct attr_check_item *ccheck = NULL;
1322
1323 if (!check) {
1324 check = attr_check_initl("crlf", "ident", "filter",
1325 "eol", "text", "working-tree-encoding",
1326 NULL);
1327 user_convert_tail = &user_convert;
1328 git_config(read_convert_config, NULL);
1329 }
1330
1331 git_check_attr(istate, path, check);
1332 ccheck = check->items;
1333 ca->crlf_action = git_path_check_crlf(ccheck + 4);
1334 if (ca->crlf_action == CRLF_UNDEFINED)
1335 ca->crlf_action = git_path_check_crlf(ccheck + 0);
1336 ca->ident = git_path_check_ident(ccheck + 1);
1337 ca->drv = git_path_check_convert(ccheck + 2);
1338 if (ca->crlf_action != CRLF_BINARY) {
1339 enum eol eol_attr = git_path_check_eol(ccheck + 3);
1340 if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_LF)
1341 ca->crlf_action = CRLF_AUTO_INPUT;
1342 else if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_CRLF)
1343 ca->crlf_action = CRLF_AUTO_CRLF;
1344 else if (eol_attr == EOL_LF)
1345 ca->crlf_action = CRLF_TEXT_INPUT;
1346 else if (eol_attr == EOL_CRLF)
1347 ca->crlf_action = CRLF_TEXT_CRLF;
1348 }
1349 ca->working_tree_encoding = git_path_check_encoding(ccheck + 5);
1350
1351 /* Save attr and make a decision for action */
1352 ca->attr_action = ca->crlf_action;
1353 if (ca->crlf_action == CRLF_TEXT)
1354 ca->crlf_action = text_eol_is_crlf() ? CRLF_TEXT_CRLF : CRLF_TEXT_INPUT;
1355 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_FALSE)
1356 ca->crlf_action = CRLF_BINARY;
1357 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_TRUE)
1358 ca->crlf_action = CRLF_AUTO_CRLF;
1359 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_INPUT)
1360 ca->crlf_action = CRLF_AUTO_INPUT;
1361 }
1362
1363 void reset_parsed_attributes(void)
1364 {
1365 struct convert_driver *drv, *next;
1366
1367 attr_check_free(check);
1368 check = NULL;
1369 reset_merge_attributes();
1370
1371 for (drv = user_convert; drv; drv = next) {
1372 next = drv->next;
1373 free((void *)drv->name);
1374 free((void *)drv->smudge);
1375 free((void *)drv->clean);
1376 free((void *)drv->process);
1377 free(drv);
1378 }
1379 user_convert = NULL;
1380 user_convert_tail = NULL;
1381 }
1382
1383 int would_convert_to_git_filter_fd(struct index_state *istate, const char *path)
1384 {
1385 struct conv_attrs ca;
1386
1387 convert_attrs(istate, &ca, path);
1388 if (!ca.drv)
1389 return 0;
1390
1391 /*
1392 * Apply a filter to an fd only if the filter is required to succeed.
1393 * We must die if the filter fails, because the original data before
1394 * filtering is not available.
1395 */
1396 if (!ca.drv->required)
1397 return 0;
1398
1399 return apply_filter(path, NULL, 0, -1, NULL, ca.drv, CAP_CLEAN, NULL, NULL);
1400 }
1401
1402 const char *get_convert_attr_ascii(struct index_state *istate, const char *path)
1403 {
1404 struct conv_attrs ca;
1405
1406 convert_attrs(istate, &ca, path);
1407 switch (ca.attr_action) {
1408 case CRLF_UNDEFINED:
1409 return "";
1410 case CRLF_BINARY:
1411 return "-text";
1412 case CRLF_TEXT:
1413 return "text";
1414 case CRLF_TEXT_INPUT:
1415 return "text eol=lf";
1416 case CRLF_TEXT_CRLF:
1417 return "text eol=crlf";
1418 case CRLF_AUTO:
1419 return "text=auto";
1420 case CRLF_AUTO_CRLF:
1421 return "text=auto eol=crlf";
1422 case CRLF_AUTO_INPUT:
1423 return "text=auto eol=lf";
1424 }
1425 return "";
1426 }
1427
1428 int convert_to_git(struct index_state *istate,
1429 const char *path, const char *src, size_t len,
1430 struct strbuf *dst, int conv_flags)
1431 {
1432 int ret = 0;
1433 struct conv_attrs ca;
1434
1435 convert_attrs(istate, &ca, path);
1436
1437 ret |= apply_filter(path, src, len, -1, dst, ca.drv, CAP_CLEAN, NULL, NULL);
1438 if (!ret && ca.drv && ca.drv->required)
1439 die(_("%s: clean filter '%s' failed"), path, ca.drv->name);
1440
1441 if (ret && dst) {
1442 src = dst->buf;
1443 len = dst->len;
1444 }
1445
1446 ret |= encode_to_git(path, src, len, dst, ca.working_tree_encoding, conv_flags);
1447 if (ret && dst) {
1448 src = dst->buf;
1449 len = dst->len;
1450 }
1451
1452 if (!(conv_flags & CONV_EOL_KEEP_CRLF)) {
1453 ret |= crlf_to_git(istate, path, src, len, dst, ca.crlf_action, conv_flags);
1454 if (ret && dst) {
1455 src = dst->buf;
1456 len = dst->len;
1457 }
1458 }
1459 return ret | ident_to_git(src, len, dst, ca.ident);
1460 }
1461
1462 void convert_to_git_filter_fd(struct index_state *istate,
1463 const char *path, int fd, struct strbuf *dst,
1464 int conv_flags)
1465 {
1466 struct conv_attrs ca;
1467 convert_attrs(istate, &ca, path);
1468
1469 assert(ca.drv);
1470
1471 if (!apply_filter(path, NULL, 0, fd, dst, ca.drv, CAP_CLEAN, NULL, NULL))
1472 die(_("%s: clean filter '%s' failed"), path, ca.drv->name);
1473
1474 encode_to_git(path, dst->buf, dst->len, dst, ca.working_tree_encoding, conv_flags);
1475 crlf_to_git(istate, path, dst->buf, dst->len, dst, ca.crlf_action, conv_flags);
1476 ident_to_git(dst->buf, dst->len, dst, ca.ident);
1477 }
1478
1479 static int convert_to_working_tree_ca_internal(const struct conv_attrs *ca,
1480 const char *path, const char *src,
1481 size_t len, struct strbuf *dst,
1482 int normalizing,
1483 const struct checkout_metadata *meta,
1484 struct delayed_checkout *dco)
1485 {
1486 int ret = 0, ret_filter = 0;
1487
1488 ret |= ident_to_worktree(src, len, dst, ca->ident);
1489 if (ret) {
1490 src = dst->buf;
1491 len = dst->len;
1492 }
1493 /*
1494 * CRLF conversion can be skipped if normalizing, unless there
1495 * is a smudge or process filter (even if the process filter doesn't
1496 * support smudge). The filters might expect CRLFs.
1497 */
1498 if ((ca->drv && (ca->drv->smudge || ca->drv->process)) || !normalizing) {
1499 ret |= crlf_to_worktree(src, len, dst, ca->crlf_action);
1500 if (ret) {
1501 src = dst->buf;
1502 len = dst->len;
1503 }
1504 }
1505
1506 ret |= encode_to_worktree(path, src, len, dst, ca->working_tree_encoding);
1507 if (ret) {
1508 src = dst->buf;
1509 len = dst->len;
1510 }
1511
1512 ret_filter = apply_filter(
1513 path, src, len, -1, dst, ca->drv, CAP_SMUDGE, meta, dco);
1514 if (!ret_filter && ca->drv && ca->drv->required)
1515 die(_("%s: smudge filter %s failed"), path, ca->drv->name);
1516
1517 return ret | ret_filter;
1518 }
1519
1520 int async_convert_to_working_tree_ca(const struct conv_attrs *ca,
1521 const char *path, const char *src,
1522 size_t len, struct strbuf *dst,
1523 const struct checkout_metadata *meta,
1524 void *dco)
1525 {
1526 return convert_to_working_tree_ca_internal(ca, path, src, len, dst, 0,
1527 meta, dco);
1528 }
1529
1530 int convert_to_working_tree_ca(const struct conv_attrs *ca,
1531 const char *path, const char *src,
1532 size_t len, struct strbuf *dst,
1533 const struct checkout_metadata *meta)
1534 {
1535 return convert_to_working_tree_ca_internal(ca, path, src, len, dst, 0,
1536 meta, NULL);
1537 }
1538
1539 int renormalize_buffer(struct index_state *istate, const char *path,
1540 const char *src, size_t len, struct strbuf *dst)
1541 {
1542 struct conv_attrs ca;
1543 int ret;
1544
1545 convert_attrs(istate, &ca, path);
1546 ret = convert_to_working_tree_ca_internal(&ca, path, src, len, dst, 1,
1547 NULL, NULL);
1548 if (ret) {
1549 src = dst->buf;
1550 len = dst->len;
1551 }
1552 return ret | convert_to_git(istate, path, src, len, dst, CONV_EOL_RENORMALIZE);
1553 }
1554
1555 /*****************************************************************
1556 *
1557 * Streaming conversion support
1558 *
1559 *****************************************************************/
1560
1561 typedef int (*filter_fn)(struct stream_filter *,
1562 const char *input, size_t *isize_p,
1563 char *output, size_t *osize_p);
1564 typedef void (*free_fn)(struct stream_filter *);
1565
1566 struct stream_filter_vtbl {
1567 filter_fn filter;
1568 free_fn free;
1569 };
1570
1571 struct stream_filter {
1572 struct stream_filter_vtbl *vtbl;
1573 };
1574
1575 static int null_filter_fn(struct stream_filter *filter UNUSED,
1576 const char *input, size_t *isize_p,
1577 char *output, size_t *osize_p)
1578 {
1579 size_t count;
1580
1581 if (!input)
1582 return 0; /* we do not keep any states */
1583 count = *isize_p;
1584 if (*osize_p < count)
1585 count = *osize_p;
1586 if (count) {
1587 memmove(output, input, count);
1588 *isize_p -= count;
1589 *osize_p -= count;
1590 }
1591 return 0;
1592 }
1593
1594 static void null_free_fn(struct stream_filter *filter UNUSED)
1595 {
1596 ; /* nothing -- null instances are shared */
1597 }
1598
1599 static struct stream_filter_vtbl null_vtbl = {
1600 .filter = null_filter_fn,
1601 .free = null_free_fn,
1602 };
1603
1604 static struct stream_filter null_filter_singleton = {
1605 .vtbl = &null_vtbl,
1606 };
1607
1608 int is_null_stream_filter(struct stream_filter *filter)
1609 {
1610 return filter == &null_filter_singleton;
1611 }
1612
1613
1614 /*
1615 * LF-to-CRLF filter
1616 */
1617
1618 struct lf_to_crlf_filter {
1619 struct stream_filter filter;
1620 unsigned has_held:1;
1621 char held;
1622 };
1623
1624 static int lf_to_crlf_filter_fn(struct stream_filter *filter,
1625 const char *input, size_t *isize_p,
1626 char *output, size_t *osize_p)
1627 {
1628 size_t count, o = 0;
1629 struct lf_to_crlf_filter *lf_to_crlf = (struct lf_to_crlf_filter *)filter;
1630
1631 /*
1632 * We may be holding onto the CR to see if it is followed by a
1633 * LF, in which case we would need to go to the main loop.
1634 * Otherwise, just emit it to the output stream.
1635 */
1636 if (lf_to_crlf->has_held && (lf_to_crlf->held != '\r' || !input)) {
1637 output[o++] = lf_to_crlf->held;
1638 lf_to_crlf->has_held = 0;
1639 }
1640
1641 /* We are told to drain */
1642 if (!input) {
1643 *osize_p -= o;
1644 return 0;
1645 }
1646
1647 count = *isize_p;
1648 if (count || lf_to_crlf->has_held) {
1649 size_t i;
1650 int was_cr = 0;
1651
1652 if (lf_to_crlf->has_held) {
1653 was_cr = 1;
1654 lf_to_crlf->has_held = 0;
1655 }
1656
1657 for (i = 0; o < *osize_p && i < count; i++) {
1658 char ch = input[i];
1659
1660 if (ch == '\n') {
1661 output[o++] = '\r';
1662 } else if (was_cr) {
1663 /*
1664 * Previous round saw CR and it is not followed
1665 * by a LF; emit the CR before processing the
1666 * current character.
1667 */
1668 output[o++] = '\r';
1669 }
1670
1671 /*
1672 * We may have consumed the last output slot,
1673 * in which case we need to break out of this
1674 * loop; hold the current character before
1675 * returning.
1676 */
1677 if (*osize_p <= o) {
1678 lf_to_crlf->has_held = 1;
1679 lf_to_crlf->held = ch;
1680 continue; /* break but increment i */
1681 }
1682
1683 if (ch == '\r') {
1684 was_cr = 1;
1685 continue;
1686 }
1687
1688 was_cr = 0;
1689 output[o++] = ch;
1690 }
1691
1692 *osize_p -= o;
1693 *isize_p -= i;
1694
1695 if (!lf_to_crlf->has_held && was_cr) {
1696 lf_to_crlf->has_held = 1;
1697 lf_to_crlf->held = '\r';
1698 }
1699 }
1700 return 0;
1701 }
1702
1703 static void lf_to_crlf_free_fn(struct stream_filter *filter)
1704 {
1705 free(filter);
1706 }
1707
1708 static struct stream_filter_vtbl lf_to_crlf_vtbl = {
1709 .filter = lf_to_crlf_filter_fn,
1710 .free = lf_to_crlf_free_fn,
1711 };
1712
1713 static struct stream_filter *lf_to_crlf_filter(void)
1714 {
1715 struct lf_to_crlf_filter *lf_to_crlf = xcalloc(1, sizeof(*lf_to_crlf));
1716
1717 lf_to_crlf->filter.vtbl = &lf_to_crlf_vtbl;
1718 return (struct stream_filter *)lf_to_crlf;
1719 }
1720
1721 /*
1722 * Cascade filter
1723 */
1724 #define FILTER_BUFFER 1024
1725 struct cascade_filter {
1726 struct stream_filter filter;
1727 struct stream_filter *one;
1728 struct stream_filter *two;
1729 char buf[FILTER_BUFFER];
1730 int end, ptr;
1731 };
1732
1733 static int cascade_filter_fn(struct stream_filter *filter,
1734 const char *input, size_t *isize_p,
1735 char *output, size_t *osize_p)
1736 {
1737 struct cascade_filter *cas = (struct cascade_filter *) filter;
1738 size_t filled = 0;
1739 size_t sz = *osize_p;
1740 size_t to_feed, remaining;
1741
1742 /*
1743 * input -- (one) --> buf -- (two) --> output
1744 */
1745 while (filled < sz) {
1746 remaining = sz - filled;
1747
1748 /* do we already have something to feed two with? */
1749 if (cas->ptr < cas->end) {
1750 to_feed = cas->end - cas->ptr;
1751 if (stream_filter(cas->two,
1752 cas->buf + cas->ptr, &to_feed,
1753 output + filled, &remaining))
1754 return -1;
1755 cas->ptr += (cas->end - cas->ptr) - to_feed;
1756 filled = sz - remaining;
1757 continue;
1758 }
1759
1760 /* feed one from upstream and have it emit into our buffer */
1761 to_feed = input ? *isize_p : 0;
1762 if (input && !to_feed)
1763 break;
1764 remaining = sizeof(cas->buf);
1765 if (stream_filter(cas->one,
1766 input, &to_feed,
1767 cas->buf, &remaining))
1768 return -1;
1769 cas->end = sizeof(cas->buf) - remaining;
1770 cas->ptr = 0;
1771 if (input) {
1772 size_t fed = *isize_p - to_feed;
1773 *isize_p -= fed;
1774 input += fed;
1775 }
1776
1777 /* do we know that we drained one completely? */
1778 if (input || cas->end)
1779 continue;
1780
1781 /* tell two to drain; we have nothing more to give it */
1782 to_feed = 0;
1783 remaining = sz - filled;
1784 if (stream_filter(cas->two,
1785 NULL, &to_feed,
1786 output + filled, &remaining))
1787 return -1;
1788 if (remaining == (sz - filled))
1789 break; /* completely drained two */
1790 filled = sz - remaining;
1791 }
1792 *osize_p -= filled;
1793 return 0;
1794 }
1795
1796 static void cascade_free_fn(struct stream_filter *filter)
1797 {
1798 struct cascade_filter *cas = (struct cascade_filter *)filter;
1799 free_stream_filter(cas->one);
1800 free_stream_filter(cas->two);
1801 free(filter);
1802 }
1803
1804 static struct stream_filter_vtbl cascade_vtbl = {
1805 .filter = cascade_filter_fn,
1806 .free = cascade_free_fn,
1807 };
1808
1809 static struct stream_filter *cascade_filter(struct stream_filter *one,
1810 struct stream_filter *two)
1811 {
1812 struct cascade_filter *cascade;
1813
1814 if (!one || is_null_stream_filter(one))
1815 return two;
1816 if (!two || is_null_stream_filter(two))
1817 return one;
1818
1819 cascade = xmalloc(sizeof(*cascade));
1820 cascade->one = one;
1821 cascade->two = two;
1822 cascade->end = cascade->ptr = 0;
1823 cascade->filter.vtbl = &cascade_vtbl;
1824 return (struct stream_filter *)cascade;
1825 }
1826
1827 /*
1828 * ident filter
1829 */
1830 #define IDENT_DRAINING (-1)
1831 #define IDENT_SKIPPING (-2)
1832 struct ident_filter {
1833 struct stream_filter filter;
1834 struct strbuf left;
1835 int state;
1836 char ident[GIT_MAX_HEXSZ + 5]; /* ": x40 $" */
1837 };
1838
1839 static int is_foreign_ident(const char *str)
1840 {
1841 int i;
1842
1843 if (!skip_prefix(str, "$Id: ", &str))
1844 return 0;
1845 for (i = 0; str[i]; i++) {
1846 if (isspace(str[i]) && str[i+1] != '$')
1847 return 1;
1848 }
1849 return 0;
1850 }
1851
1852 static void ident_drain(struct ident_filter *ident, char **output_p, size_t *osize_p)
1853 {
1854 size_t to_drain = ident->left.len;
1855
1856 if (*osize_p < to_drain)
1857 to_drain = *osize_p;
1858 if (to_drain) {
1859 memcpy(*output_p, ident->left.buf, to_drain);
1860 strbuf_remove(&ident->left, 0, to_drain);
1861 *output_p += to_drain;
1862 *osize_p -= to_drain;
1863 }
1864 if (!ident->left.len)
1865 ident->state = 0;
1866 }
1867
1868 static int ident_filter_fn(struct stream_filter *filter,
1869 const char *input, size_t *isize_p,
1870 char *output, size_t *osize_p)
1871 {
1872 struct ident_filter *ident = (struct ident_filter *)filter;
1873 static const char head[] = "$Id";
1874
1875 if (!input) {
1876 /* drain upon eof */
1877 switch (ident->state) {
1878 default:
1879 strbuf_add(&ident->left, head, ident->state);
1880 /* fallthrough */
1881 case IDENT_SKIPPING:
1882 /* fallthrough */
1883 case IDENT_DRAINING:
1884 ident_drain(ident, &output, osize_p);
1885 }
1886 return 0;
1887 }
1888
1889 while (*isize_p || (ident->state == IDENT_DRAINING)) {
1890 int ch;
1891
1892 if (ident->state == IDENT_DRAINING) {
1893 ident_drain(ident, &output, osize_p);
1894 if (!*osize_p)
1895 break;
1896 continue;
1897 }
1898
1899 ch = *(input++);
1900 (*isize_p)--;
1901
1902 if (ident->state == IDENT_SKIPPING) {
1903 /*
1904 * Skipping until '$' or LF, but keeping them
1905 * in case it is a foreign ident.
1906 */
1907 strbuf_addch(&ident->left, ch);
1908 if (ch != '\n' && ch != '$')
1909 continue;
1910 if (ch == '$' && !is_foreign_ident(ident->left.buf)) {
1911 strbuf_setlen(&ident->left, sizeof(head) - 1);
1912 strbuf_addstr(&ident->left, ident->ident);
1913 }
1914 ident->state = IDENT_DRAINING;
1915 continue;
1916 }
1917
1918 if (ident->state < sizeof(head) &&
1919 head[ident->state] == ch) {
1920 ident->state++;
1921 continue;
1922 }
1923
1924 if (ident->state)
1925 strbuf_add(&ident->left, head, ident->state);
1926 if (ident->state == sizeof(head) - 1) {
1927 if (ch != ':' && ch != '$') {
1928 strbuf_addch(&ident->left, ch);
1929 ident->state = 0;
1930 continue;
1931 }
1932
1933 if (ch == ':') {
1934 strbuf_addch(&ident->left, ch);
1935 ident->state = IDENT_SKIPPING;
1936 } else {
1937 strbuf_addstr(&ident->left, ident->ident);
1938 ident->state = IDENT_DRAINING;
1939 }
1940 continue;
1941 }
1942
1943 strbuf_addch(&ident->left, ch);
1944 ident->state = IDENT_DRAINING;
1945 }
1946 return 0;
1947 }
1948
1949 static void ident_free_fn(struct stream_filter *filter)
1950 {
1951 struct ident_filter *ident = (struct ident_filter *)filter;
1952 strbuf_release(&ident->left);
1953 free(filter);
1954 }
1955
1956 static struct stream_filter_vtbl ident_vtbl = {
1957 .filter = ident_filter_fn,
1958 .free = ident_free_fn,
1959 };
1960
1961 static struct stream_filter *ident_filter(const struct object_id *oid)
1962 {
1963 struct ident_filter *ident = xmalloc(sizeof(*ident));
1964
1965 xsnprintf(ident->ident, sizeof(ident->ident),
1966 ": %s $", oid_to_hex(oid));
1967 strbuf_init(&ident->left, 0);
1968 ident->filter.vtbl = &ident_vtbl;
1969 ident->state = 0;
1970 return (struct stream_filter *)ident;
1971 }
1972
1973 /*
1974 * Return an appropriately constructed filter for the given ca, or NULL if
1975 * the contents cannot be filtered without reading the whole thing
1976 * in-core.
1977 *
1978 * Note that you would be crazy to set CRLF, smudge/clean or ident to a
1979 * large binary blob you would want us not to slurp into the memory!
1980 */
1981 struct stream_filter *get_stream_filter_ca(const struct conv_attrs *ca,
1982 const struct object_id *oid)
1983 {
1984 struct stream_filter *filter = NULL;
1985
1986 if (classify_conv_attrs(ca) != CA_CLASS_STREAMABLE)
1987 return NULL;
1988
1989 if (ca->ident)
1990 filter = ident_filter(oid);
1991
1992 if (output_eol(ca->crlf_action) == EOL_CRLF)
1993 filter = cascade_filter(filter, lf_to_crlf_filter());
1994 else
1995 filter = cascade_filter(filter, &null_filter_singleton);
1996
1997 return filter;
1998 }
1999
2000 struct stream_filter *get_stream_filter(struct index_state *istate,
2001 const char *path,
2002 const struct object_id *oid)
2003 {
2004 struct conv_attrs ca;
2005 convert_attrs(istate, &ca, path);
2006 return get_stream_filter_ca(&ca, oid);
2007 }
2008
2009 void free_stream_filter(struct stream_filter *filter)
2010 {
2011 filter->vtbl->free(filter);
2012 }
2013
2014 int stream_filter(struct stream_filter *filter,
2015 const char *input, size_t *isize_p,
2016 char *output, size_t *osize_p)
2017 {
2018 return filter->vtbl->filter(filter, input, isize_p, output, osize_p);
2019 }
2020
2021 void init_checkout_metadata(struct checkout_metadata *meta, const char *refname,
2022 const struct object_id *treeish,
2023 const struct object_id *blob)
2024 {
2025 memset(meta, 0, sizeof(*meta));
2026 if (refname)
2027 meta->refname = refname;
2028 if (treeish)
2029 oidcpy(&meta->treeish, treeish);
2030 if (blob)
2031 oidcpy(&meta->blob, blob);
2032 }
2033
2034 void clone_checkout_metadata(struct checkout_metadata *dst,
2035 const struct checkout_metadata *src,
2036 const struct object_id *blob)
2037 {
2038 memcpy(dst, src, sizeof(*dst));
2039 if (blob)
2040 oidcpy(&dst->blob, blob);
2041 }
2042
2043 enum conv_attrs_classification classify_conv_attrs(const struct conv_attrs *ca)
2044 {
2045 if (ca->drv) {
2046 if (ca->drv->process)
2047 return CA_CLASS_INCORE_PROCESS;
2048 if (ca->drv->smudge || ca->drv->clean)
2049 return CA_CLASS_INCORE_FILTER;
2050 }
2051
2052 if (ca->working_tree_encoding)
2053 return CA_CLASS_INCORE;
2054
2055 if (ca->crlf_action == CRLF_AUTO || ca->crlf_action == CRLF_AUTO_CRLF)
2056 return CA_CLASS_INCORE;
2057
2058 return CA_CLASS_STREAMABLE;
2059 }
The Best VCS Around