?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge branch 'jk/bundle-use-dash-for-stdfiles'
[git/debian.git] / convert.c
blob349c7e4af155357417b937836da43c84111435b8
1 #include "cache.h"
2 #include "config.h"
3 #include "hex.h"
4 #include "object-store.h"
5 #include "attr.h"
6 #include "run-command.h"
7 #include "quote.h"
8 #include "sigchain.h"
9 #include "pkt-line.h"
10 #include "sub-process.h"
11 #include "utf8.h"
12 #include "ll-merge.h"
13
14 /*
15 * convert.c - convert a file when checking it out and checking it in.
16 *
17 * This should use the pathname to decide on whether it wants to do some
18 * more interesting conversions (automatic gzip/unzip, general format
19 * conversions etc etc), but by default it just does automatic CRLF<->LF
20 * translation when the "text" attribute or "auto_crlf" option is set.
21 */
22
23 /* Stat bits: When BIN is set, the txt bits are unset */
24 #define CONVERT_STAT_BITS_TXT_LF 0x1
25 #define CONVERT_STAT_BITS_TXT_CRLF 0x2
26 #define CONVERT_STAT_BITS_BIN 0x4
27
28 struct text_stat {
29 /* NUL, CR, LF and CRLF counts */
30 unsigned nul, lonecr, lonelf, crlf;
31
32 /* These are just approximations! */
33 unsigned printable, nonprintable;
34 };
35
36 static void gather_stats(const char *buf, unsigned long size, struct text_stat *stats)
37 {
38 unsigned long i;
39
40 memset(stats, 0, sizeof(*stats));
41
42 for (i = 0; i < size; i++) {
43 unsigned char c = buf[i];
44 if (c == '\r') {
45 if (i+1 < size && buf[i+1] == '\n') {
46 stats->crlf++;
47 i++;
48 } else
49 stats->lonecr++;
50 continue;
51 }
52 if (c == '\n') {
53 stats->lonelf++;
54 continue;
55 }
56 if (c == 127)
57 /* DEL */
58 stats->nonprintable++;
59 else if (c < 32) {
60 switch (c) {
61 /* BS, HT, ESC and FF */
62 case '\b': case '\t': case '\033': case '\014':
63 stats->printable++;
64 break;
65 case 0:
66 stats->nul++;
67 /* fall through */
68 default:
69 stats->nonprintable++;
70 }
71 }
72 else
73 stats->printable++;
74 }
75
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--;
79 }
80
81 /*
82 * The same heuristics as diff.c::mmfile_is_binary()
83 * We treat files with bare CR as binary
84 */
85 static int convert_is_binary(const struct text_stat *stats)
86 {
87 if (stats->lonecr)
88 return 1;
89 if (stats->nul)
90 return 1;
91 if ((stats->printable >> 7) < stats->nonprintable)
92 return 1;
93 return 0;
94 }
95
96 static unsigned int gather_convert_stats(const char *data, unsigned long size)
97 {
98 struct text_stat stats;
99 int ret = 0;
100 if (!data || !size)
101 return 0;
102 gather_stats(data, size, &stats);
103 if (convert_is_binary(&stats))
104 ret |= CONVERT_STAT_BITS_BIN;
105 if (stats.crlf)
106 ret |= CONVERT_STAT_BITS_TXT_CRLF;
107 if (stats.lonelf)
108 ret |= CONVERT_STAT_BITS_TXT_LF;
109
110 return ret;
111 }
112
113 static const char *gather_convert_stats_ascii(const char *data, unsigned long size)
114 {
115 unsigned int convert_stats = gather_convert_stats(data, size);
116
117 if (convert_stats & CONVERT_STAT_BITS_BIN)
118 return "-text";
119 switch (convert_stats) {
120 case CONVERT_STAT_BITS_TXT_LF:
121 return "lf";
122 case CONVERT_STAT_BITS_TXT_CRLF:
123 return "crlf";
124 case CONVERT_STAT_BITS_TXT_LF | CONVERT_STAT_BITS_TXT_CRLF:
125 return "mixed";
126 default:
127 return "none";
128 }
129 }
130
131 const char *get_cached_convert_stats_ascii(struct index_state *istate,
132 const char *path)
133 {
134 const char *ret;
135 unsigned long sz;
136 void *data = read_blob_data_from_index(istate, path, &sz);
137 ret = gather_convert_stats_ascii(data, sz);
138 free(data);
139 return ret;
140 }
141
142 const char *get_wt_convert_stats_ascii(const char *path)
143 {
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);
148 strbuf_release(&sb);
149 return ret;
150 }
151
152 static int text_eol_is_crlf(void)
153 {
154 if (auto_crlf == AUTO_CRLF_TRUE)
155 return 1;
156 else if (auto_crlf == AUTO_CRLF_INPUT)
157 return 0;
158 if (core_eol == EOL_CRLF)
159 return 1;
160 if (core_eol == EOL_UNSET && EOL_NATIVE == EOL_CRLF)
161 return 1;
162 return 0;
163 }
164
165 static enum eol output_eol(enum convert_crlf_action crlf_action)
166 {
167 switch (crlf_action) {
168 case CRLF_BINARY:
169 return EOL_UNSET;
170 case CRLF_TEXT_CRLF:
171 return EOL_CRLF;
172 case CRLF_TEXT_INPUT:
173 return EOL_LF;
174 case CRLF_UNDEFINED:
175 case CRLF_AUTO_CRLF:
176 return EOL_CRLF;
177 case CRLF_AUTO_INPUT:
178 return EOL_LF;
179 case CRLF_TEXT:
180 case CRLF_AUTO:
181 /* fall through */
182 return text_eol_is_crlf() ? EOL_CRLF : EOL_LF;
183 }
184 warning(_("illegal crlf_action %d"), (int)crlf_action);
185 return core_eol;
186 }
187
188 static void check_global_conv_flags_eol(const char *path,
189 struct text_stat *old_stats, struct text_stat *new_stats,
190 int conv_flags)
191 {
192 if (old_stats->crlf && !new_stats->crlf ) {
193 /*
194 * CRLFs would not be restored by checkout
195 */
196 if (conv_flags & CONV_EOL_RNDTRP_DIE)
197 die(_("CRLF would be replaced by LF in %s"), path);
198 else if (conv_flags & CONV_EOL_RNDTRP_WARN)
199 warning(_("in the working copy of '%s', CRLF will be"
200 " replaced by LF the next time Git touches"
201 " it"), path);
202 } else if (old_stats->lonelf && !new_stats->lonelf ) {
203 /*
204 * CRLFs would be added by checkout
205 */
206 if (conv_flags & CONV_EOL_RNDTRP_DIE)
207 die(_("LF would be replaced by CRLF in %s"), path);
208 else if (conv_flags & CONV_EOL_RNDTRP_WARN)
209 warning(_("in the working copy of '%s', LF will be"
210 " replaced by CRLF the next time Git touches"
211 " it"), path);
212 }
213 }
214
215 static int has_crlf_in_index(struct index_state *istate, const char *path)
216 {
217 unsigned long sz;
218 void *data;
219 const char *crp;
220 int has_crlf = 0;
221
222 data = read_blob_data_from_index(istate, path, &sz);
223 if (!data)
224 return 0;
225
226 crp = memchr(data, '\r', sz);
227 if (crp) {
228 unsigned int ret_stats;
229 ret_stats = gather_convert_stats(data, sz);
230 if (!(ret_stats & CONVERT_STAT_BITS_BIN) &&
231 (ret_stats & CONVERT_STAT_BITS_TXT_CRLF))
232 has_crlf = 1;
233 }
234 free(data);
235 return has_crlf;
236 }
237
238 static int will_convert_lf_to_crlf(struct text_stat *stats,
239 enum convert_crlf_action crlf_action)
240 {
241 if (output_eol(crlf_action) != EOL_CRLF)
242 return 0;
243 /* No "naked" LF? Nothing to convert, regardless. */
244 if (!stats->lonelf)
245 return 0;
246
247 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
248 /* If we have any CR or CRLF line endings, we do not touch it */
249 /* This is the new safer autocrlf-handling */
250 if (stats->lonecr || stats->crlf)
251 return 0;
252
253 if (convert_is_binary(stats))
254 return 0;
255 }
256 return 1;
257
258 }
259
260 static int validate_encoding(const char *path, const char *enc,
261 const char *data, size_t len, int die_on_error)
262 {
263 const char *stripped;
264
265 /* We only check for UTF here as UTF?? can be an alias for UTF-?? */
266 if (skip_iprefix(enc, "UTF", &stripped)) {
267 skip_prefix(stripped, "-", &stripped);
268
269 /*
270 * Check for detectable errors in UTF encodings
271 */
272 if (has_prohibited_utf_bom(enc, data, len)) {
273 const char *error_msg = _(
274 "BOM is prohibited in '%s' if encoded as %s");
275 /*
276 * This advice is shown for UTF-??BE and UTF-??LE encodings.
277 * We cut off the last two characters of the encoding name
278 * to generate the encoding name suitable for BOMs.
279 */
280 const char *advise_msg = _(
281 "The file '%s' contains a byte order "
282 "mark (BOM). Please use UTF-%.*s as "
283 "working-tree-encoding.");
284 int stripped_len = strlen(stripped) - strlen("BE");
285 advise(advise_msg, path, stripped_len, stripped);
286 if (die_on_error)
287 die(error_msg, path, enc);
288 else {
289 return error(error_msg, path, enc);
290 }
291
292 } else if (is_missing_required_utf_bom(enc, data, len)) {
293 const char *error_msg = _(
294 "BOM is required in '%s' if encoded as %s");
295 const char *advise_msg = _(
296 "The file '%s' is missing a byte order "
297 "mark (BOM). Please use UTF-%sBE or UTF-%sLE "
298 "(depending on the byte order) as "
299 "working-tree-encoding.");
300 advise(advise_msg, path, stripped, stripped);
301 if (die_on_error)
302 die(error_msg, path, enc);
303 else {
304 return error(error_msg, path, enc);
305 }
306 }
307
308 }
309 return 0;
310 }
311
312 static void trace_encoding(const char *context, const char *path,
313 const char *encoding, const char *buf, size_t len)
314 {
315 static struct trace_key coe = TRACE_KEY_INIT(WORKING_TREE_ENCODING);
316 struct strbuf trace = STRBUF_INIT;
317 int i;
318
319 strbuf_addf(&trace, "%s (%s, considered %s):\n", context, path, encoding);
320 for (i = 0; i < len && buf; ++i) {
321 strbuf_addf(
322 &trace, "| \033[2m%2i:\033[0m %2x \033[2m%c\033[0m%c",
323 i,
324 (unsigned char) buf[i],
325 (buf[i] > 32 && buf[i] < 127 ? buf[i] : ' '),
326 ((i+1) % 8 && (i+1) < len ? ' ' : '\n')
327 );
328 }
329 strbuf_addchars(&trace, '\n', 1);
330
331 trace_strbuf(&coe, &trace);
332 strbuf_release(&trace);
333 }
334
335 static int check_roundtrip(const char *enc_name)
336 {
337 /*
338 * check_roundtrip_encoding contains a string of comma and/or
339 * space separated encodings (eg. "UTF-16, ASCII, CP1125").
340 * Search for the given encoding in that string.
341 */
342 const char *found = strcasestr(check_roundtrip_encoding, enc_name);
343 const char *next;
344 int len;
345 if (!found)
346 return 0;
347 next = found + strlen(enc_name);
348 len = strlen(check_roundtrip_encoding);
349 return (found && (
350 /*
351 * check that the found encoding is at the
352 * beginning of check_roundtrip_encoding or
353 * that it is prefixed with a space or comma
354 */
355 found == check_roundtrip_encoding || (
356 (isspace(found[-1]) || found[-1] == ',')
357 )
358 ) && (
359 /*
360 * check that the found encoding is at the
361 * end of check_roundtrip_encoding or
362 * that it is suffixed with a space or comma
363 */
364 next == check_roundtrip_encoding + len || (
365 next < check_roundtrip_encoding + len &&
366 (isspace(next[0]) || next[0] == ',')
367 )
368 ));
369 }
370
371 static const char *default_encoding = "UTF-8";
372
373 static int encode_to_git(const char *path, const char *src, size_t src_len,
374 struct strbuf *buf, const char *enc, int conv_flags)
375 {
376 char *dst;
377 size_t dst_len;
378 int die_on_error = conv_flags & CONV_WRITE_OBJECT;
379
380 /*
381 * No encoding is specified or there is nothing to encode.
382 * Tell the caller that the content was not modified.
383 */
384 if (!enc || (src && !src_len))
385 return 0;
386
387 /*
388 * Looks like we got called from "would_convert_to_git()".
389 * This means Git wants to know if it would encode (= modify!)
390 * the content. Let's answer with "yes", since an encoding was
391 * specified.
392 */
393 if (!buf && !src)
394 return 1;
395
396 if (validate_encoding(path, enc, src, src_len, die_on_error))
397 return 0;
398
399 trace_encoding("source", path, enc, src, src_len);
400 dst = reencode_string_len(src, src_len, default_encoding, enc,
401 &dst_len);
402 if (!dst) {
403 /*
404 * We could add the blob "as-is" to Git. However, on checkout
405 * we would try to re-encode to the original encoding. This
406 * would fail and we would leave the user with a messed-up
407 * working tree. Let's try to avoid this by screaming loud.
408 */
409 const char* msg = _("failed to encode '%s' from %s to %s");
410 if (die_on_error)
411 die(msg, path, enc, default_encoding);
412 else {
413 error(msg, path, enc, default_encoding);
414 return 0;
415 }
416 }
417 trace_encoding("destination", path, default_encoding, dst, dst_len);
418
419 /*
420 * UTF supports lossless conversion round tripping [1] and conversions
421 * between UTF and other encodings are mostly round trip safe as
422 * Unicode aims to be a superset of all other character encodings.
423 * However, certain encodings (e.g. SHIFT-JIS) are known to have round
424 * trip issues [2]. Check the round trip conversion for all encodings
425 * listed in core.checkRoundtripEncoding.
426 *
427 * The round trip check is only performed if content is written to Git.
428 * This ensures that no information is lost during conversion to/from
429 * the internal UTF-8 representation.
430 *
431 * Please note, the code below is not tested because I was not able to
432 * generate a faulty round trip without an iconv error. Iconv errors
433 * are already caught above.
434 *
435 * [1] http://unicode.org/faq/utf_bom.html#gen2
436 * [2] https://support.microsoft.com/en-us/help/170559/prb-conversion-problem-between-shift-jis-and-unicode
437 */
438 if (die_on_error && check_roundtrip(enc)) {
439 char *re_src;
440 size_t re_src_len;
441
442 re_src = reencode_string_len(dst, dst_len,
443 enc, default_encoding,
444 &re_src_len);
445
446 trace_printf("Checking roundtrip encoding for %s...\n", enc);
447 trace_encoding("reencoded source", path, enc,
448 re_src, re_src_len);
449
450 if (!re_src || src_len != re_src_len ||
451 memcmp(src, re_src, src_len)) {
452 const char* msg = _("encoding '%s' from %s to %s and "
453 "back is not the same");
454 die(msg, path, enc, default_encoding);
455 }
456
457 free(re_src);
458 }
459
460 strbuf_attach(buf, dst, dst_len, dst_len + 1);
461 return 1;
462 }
463
464 static int encode_to_worktree(const char *path, const char *src, size_t src_len,
465 struct strbuf *buf, const char *enc)
466 {
467 char *dst;
468 size_t dst_len;
469
470 /*
471 * No encoding is specified or there is nothing to encode.
472 * Tell the caller that the content was not modified.
473 */
474 if (!enc || (src && !src_len))
475 return 0;
476
477 dst = reencode_string_len(src, src_len, enc, default_encoding,
478 &dst_len);
479 if (!dst) {
480 error(_("failed to encode '%s' from %s to %s"),
481 path, default_encoding, enc);
482 return 0;
483 }
484
485 strbuf_attach(buf, dst, dst_len, dst_len + 1);
486 return 1;
487 }
488
489 static int crlf_to_git(struct index_state *istate,
490 const char *path, const char *src, size_t len,
491 struct strbuf *buf,
492 enum convert_crlf_action crlf_action, int conv_flags)
493 {
494 struct text_stat stats;
495 char *dst;
496 int convert_crlf_into_lf;
497
498 if (crlf_action == CRLF_BINARY ||
499 (src && !len))
500 return 0;
501
502 /*
503 * If we are doing a dry-run and have no source buffer, there is
504 * nothing to analyze; we must assume we would convert.
505 */
506 if (!buf && !src)
507 return 1;
508
509 gather_stats(src, len, &stats);
510 /* Optimization: No CRLF? Nothing to convert, regardless. */
511 convert_crlf_into_lf = !!stats.crlf;
512
513 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
514 if (convert_is_binary(&stats))
515 return 0;
516 /*
517 * If the file in the index has any CR in it, do not
518 * convert. This is the new safer autocrlf handling,
519 * unless we want to renormalize in a merge or
520 * cherry-pick.
521 */
522 if ((!(conv_flags & CONV_EOL_RENORMALIZE)) &&
523 has_crlf_in_index(istate, path))
524 convert_crlf_into_lf = 0;
525 }
526 if (((conv_flags & CONV_EOL_RNDTRP_WARN) ||
527 ((conv_flags & CONV_EOL_RNDTRP_DIE) && len))) {
528 struct text_stat new_stats;
529 memcpy(&new_stats, &stats, sizeof(new_stats));
530 /* simulate "git add" */
531 if (convert_crlf_into_lf) {
532 new_stats.lonelf += new_stats.crlf;
533 new_stats.crlf = 0;
534 }
535 /* simulate "git checkout" */
536 if (will_convert_lf_to_crlf(&new_stats, crlf_action)) {
537 new_stats.crlf += new_stats.lonelf;
538 new_stats.lonelf = 0;
539 }
540 check_global_conv_flags_eol(path, &stats, &new_stats, conv_flags);
541 }
542 if (!convert_crlf_into_lf)
543 return 0;
544
545 /*
546 * At this point all of our source analysis is done, and we are sure we
547 * would convert. If we are in dry-run mode, we can give an answer.
548 */
549 if (!buf)
550 return 1;
551
552 /* only grow if not in place */
553 if (strbuf_avail(buf) + buf->len < len)
554 strbuf_grow(buf, len - buf->len);
555 dst = buf->buf;
556 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
557 /*
558 * If we guessed, we already know we rejected a file with
559 * lone CR, and we can strip a CR without looking at what
560 * follow it.
561 */
562 do {
563 unsigned char c = *src++;
564 if (c != '\r')
565 *dst++ = c;
566 } while (--len);
567 } else {
568 do {
569 unsigned char c = *src++;
570 if (! (c == '\r' && (1 < len && *src == '\n')))
571 *dst++ = c;
572 } while (--len);
573 }
574 strbuf_setlen(buf, dst - buf->buf);
575 return 1;
576 }
577
578 static int crlf_to_worktree(const char *src, size_t len, struct strbuf *buf,
579 enum convert_crlf_action crlf_action)
580 {
581 char *to_free = NULL;
582 struct text_stat stats;
583
584 if (!len || output_eol(crlf_action) != EOL_CRLF)
585 return 0;
586
587 gather_stats(src, len, &stats);
588 if (!will_convert_lf_to_crlf(&stats, crlf_action))
589 return 0;
590
591 /* are we "faking" in place editing ? */
592 if (src == buf->buf)
593 to_free = strbuf_detach(buf, NULL);
594
595 strbuf_grow(buf, len + stats.lonelf);
596 for (;;) {
597 const char *nl = memchr(src, '\n', len);
598 if (!nl)
599 break;
600 if (nl > src && nl[-1] == '\r') {
601 strbuf_add(buf, src, nl + 1 - src);
602 } else {
603 strbuf_add(buf, src, nl - src);
604 strbuf_addstr(buf, "\r\n");
605 }
606 len -= nl + 1 - src;
607 src = nl + 1;
608 }
609 strbuf_add(buf, src, len);
610
611 free(to_free);
612 return 1;
613 }
614
615 struct filter_params {
616 const char *src;
617 size_t size;
618 int fd;
619 const char *cmd;
620 const char *path;
621 };
622
623 static int filter_buffer_or_fd(int in UNUSED, int out, void *data)
624 {
625 /*
626 * Spawn cmd and feed the buffer contents through its stdin.
627 */
628 struct child_process child_process = CHILD_PROCESS_INIT;
629 struct filter_params *params = (struct filter_params *)data;
630 int write_err, status;
631
632 /* apply % substitution to cmd */
633 struct strbuf cmd = STRBUF_INIT;
634 struct strbuf path = STRBUF_INIT;
635 struct strbuf_expand_dict_entry dict[] = {
636 { "f", NULL, },
637 { NULL, NULL, },
638 };
639
640 /* quote the path to preserve spaces, etc. */
641 sq_quote_buf(&path, params->path);
642 dict[0].value = path.buf;
643
644 /* expand all %f with the quoted path */
645 strbuf_expand(&cmd, params->cmd, strbuf_expand_dict_cb, &dict);
646 strbuf_release(&path);
647
648 strvec_push(&child_process.args, cmd.buf);
649 child_process.use_shell = 1;
650 child_process.in = -1;
651 child_process.out = out;
652
653 if (start_command(&child_process)) {
654 strbuf_release(&cmd);
655 return error(_("cannot fork to run external filter '%s'"),
656 params->cmd);
657 }
658
659 sigchain_push(SIGPIPE, SIG_IGN);
660
661 if (params->src) {
662 write_err = (write_in_full(child_process.in,
663 params->src, params->size) < 0);
664 if (errno == EPIPE)
665 write_err = 0;
666 } else {
667 write_err = copy_fd(params->fd, child_process.in);
668 if (write_err == COPY_WRITE_ERROR && errno == EPIPE)
669 write_err = 0;
670 }
671
672 if (close(child_process.in))
673 write_err = 1;
674 if (write_err)
675 error(_("cannot feed the input to external filter '%s'"),
676 params->cmd);
677
678 sigchain_pop(SIGPIPE);
679
680 status = finish_command(&child_process);
681 if (status)
682 error(_("external filter '%s' failed %d"), params->cmd, status);
683
684 strbuf_release(&cmd);
685 return (write_err || status);
686 }
687
688 static int apply_single_file_filter(const char *path, const char *src, size_t len, int fd,
689 struct strbuf *dst, const char *cmd)
690 {
691 /*
692 * Create a pipeline to have the command filter the buffer's
693 * contents.
694 *
695 * (child --> cmd) --> us
696 */
697 int err = 0;
698 struct strbuf nbuf = STRBUF_INIT;
699 struct async async;
700 struct filter_params params;
701
702 memset(&async, 0, sizeof(async));
703 async.proc = filter_buffer_or_fd;
704 async.data = &params;
705 async.out = -1;
706 params.src = src;
707 params.size = len;
708 params.fd = fd;
709 params.cmd = cmd;
710 params.path = path;
711
712 fflush(NULL);
713 if (start_async(&async))
714 return 0; /* error was already reported */
715
716 if (strbuf_read(&nbuf, async.out, 0) < 0) {
717 err = error(_("read from external filter '%s' failed"), cmd);
718 }
719 if (close(async.out)) {
720 err = error(_("read from external filter '%s' failed"), cmd);
721 }
722 if (finish_async(&async)) {
723 err = error(_("external filter '%s' failed"), cmd);
724 }
725
726 if (!err) {
727 strbuf_swap(dst, &nbuf);
728 }
729 strbuf_release(&nbuf);
730 return !err;
731 }
732
733 #define CAP_CLEAN (1u<<0)
734 #define CAP_SMUDGE (1u<<1)
735 #define CAP_DELAY (1u<<2)
736
737 struct cmd2process {
738 struct subprocess_entry subprocess; /* must be the first member! */
739 unsigned int supported_capabilities;
740 };
741
742 static int subprocess_map_initialized;
743 static struct hashmap subprocess_map;
744
745 static int start_multi_file_filter_fn(struct subprocess_entry *subprocess)
746 {
747 static int versions[] = {2, 0};
748 static struct subprocess_capability capabilities[] = {
749 { "clean", CAP_CLEAN },
750 { "smudge", CAP_SMUDGE },
751 { "delay", CAP_DELAY },
752 { NULL, 0 }
753 };
754 struct cmd2process *entry = (struct cmd2process *)subprocess;
755 return subprocess_handshake(subprocess, "git-filter", versions, NULL,
756 capabilities,
757 &entry->supported_capabilities);
758 }
759
760 static void handle_filter_error(const struct strbuf *filter_status,
761 struct cmd2process *entry,
762 const unsigned int wanted_capability)
763 {
764 if (!strcmp(filter_status->buf, "error"))
765 ; /* The filter signaled a problem with the file. */
766 else if (!strcmp(filter_status->buf, "abort") && wanted_capability) {
767 /*
768 * The filter signaled a permanent problem. Don't try to filter
769 * files with the same command for the lifetime of the current
770 * Git process.
771 */
772 entry->supported_capabilities &= ~wanted_capability;
773 } else {
774 /*
775 * Something went wrong with the protocol filter.
776 * Force shutdown and restart if another blob requires filtering.
777 */
778 error(_("external filter '%s' failed"), entry->subprocess.cmd);
779 subprocess_stop(&subprocess_map, &entry->subprocess);
780 free(entry);
781 }
782 }
783
784 static int apply_multi_file_filter(const char *path, const char *src, size_t len,
785 int fd, struct strbuf *dst, const char *cmd,
786 const unsigned int wanted_capability,
787 const struct checkout_metadata *meta,
788 struct delayed_checkout *dco)
789 {
790 int err;
791 int can_delay = 0;
792 struct cmd2process *entry;
793 struct child_process *process;
794 struct strbuf nbuf = STRBUF_INIT;
795 struct strbuf filter_status = STRBUF_INIT;
796 const char *filter_type;
797
798 if (!subprocess_map_initialized) {
799 subprocess_map_initialized = 1;
800 hashmap_init(&subprocess_map, cmd2process_cmp, NULL, 0);
801 entry = NULL;
802 } else {
803 entry = (struct cmd2process *)subprocess_find_entry(&subprocess_map, cmd);
804 }
805
806 fflush(NULL);
807
808 if (!entry) {
809 entry = xmalloc(sizeof(*entry));
810 entry->supported_capabilities = 0;
811
812 if (subprocess_start(&subprocess_map, &entry->subprocess, cmd, start_multi_file_filter_fn)) {
813 free(entry);
814 return 0;
815 }
816 }
817 process = &entry->subprocess.process;
818
819 if (!(entry->supported_capabilities & wanted_capability))
820 return 0;
821
822 if (wanted_capability & CAP_CLEAN)
823 filter_type = "clean";
824 else if (wanted_capability & CAP_SMUDGE)
825 filter_type = "smudge";
826 else
827 die(_("unexpected filter type"));
828
829 sigchain_push(SIGPIPE, SIG_IGN);
830
831 assert(strlen(filter_type) < LARGE_PACKET_DATA_MAX - strlen("command=\n"));
832 err = packet_write_fmt_gently(process->in, "command=%s\n", filter_type);
833 if (err)
834 goto done;
835
836 err = strlen(path) > LARGE_PACKET_DATA_MAX - strlen("pathname=\n");
837 if (err) {
838 error(_("path name too long for external filter"));
839 goto done;
840 }
841
842 err = packet_write_fmt_gently(process->in, "pathname=%s\n", path);
843 if (err)
844 goto done;
845
846 if (meta && meta->refname) {
847 err = packet_write_fmt_gently(process->in, "ref=%s\n", meta->refname);
848 if (err)
849 goto done;
850 }
851
852 if (meta && !is_null_oid(&meta->treeish)) {
853 err = packet_write_fmt_gently(process->in, "treeish=%s\n", oid_to_hex(&meta->treeish));
854 if (err)
855 goto done;
856 }
857
858 if (meta && !is_null_oid(&meta->blob)) {
859 err = packet_write_fmt_gently(process->in, "blob=%s\n", oid_to_hex(&meta->blob));
860 if (err)
861 goto done;
862 }
863
864 if ((entry->supported_capabilities & CAP_DELAY) &&
865 dco && dco->state == CE_CAN_DELAY) {
866 can_delay = 1;
867 err = packet_write_fmt_gently(process->in, "can-delay=1\n");
868 if (err)
869 goto done;
870 }
871
872 err = packet_flush_gently(process->in);
873 if (err)
874 goto done;
875
876 if (fd >= 0)
877 err = write_packetized_from_fd_no_flush(fd, process->in);
878 else
879 err = write_packetized_from_buf_no_flush(src, len, process->in);
880 if (err)
881 goto done;
882
883 err = packet_flush_gently(process->in);
884 if (err)
885 goto done;
886
887 err = subprocess_read_status(process->out, &filter_status);
888 if (err)
889 goto done;
890
891 if (can_delay && !strcmp(filter_status.buf, "delayed")) {
892 string_list_insert(&dco->filters, cmd);
893 string_list_insert(&dco->paths, path);
894 } else {
895 /* The filter got the blob and wants to send us a response. */
896 err = strcmp(filter_status.buf, "success");
897 if (err)
898 goto done;
899
900 err = read_packetized_to_strbuf(process->out, &nbuf,
901 PACKET_READ_GENTLE_ON_EOF) < 0;
902 if (err)
903 goto done;
904
905 err = subprocess_read_status(process->out, &filter_status);
906 if (err)
907 goto done;
908
909 err = strcmp(filter_status.buf, "success");
910 }
911
912 done:
913 sigchain_pop(SIGPIPE);
914
915 if (err)
916 handle_filter_error(&filter_status, entry, wanted_capability);
917 else
918 strbuf_swap(dst, &nbuf);
919 strbuf_release(&nbuf);
920 strbuf_release(&filter_status);
921 return !err;
922 }
923
924
925 int async_query_available_blobs(const char *cmd, struct string_list *available_paths)
926 {
927 int err;
928 char *line;
929 struct cmd2process *entry;
930 struct child_process *process;
931 struct strbuf filter_status = STRBUF_INIT;
932
933 assert(subprocess_map_initialized);
934 entry = (struct cmd2process *)subprocess_find_entry(&subprocess_map, cmd);
935 if (!entry) {
936 error(_("external filter '%s' is not available anymore although "
937 "not all paths have been filtered"), cmd);
938 return 0;
939 }
940 process = &entry->subprocess.process;
941 sigchain_push(SIGPIPE, SIG_IGN);
942
943 err = packet_write_fmt_gently(
944 process->in, "command=list_available_blobs\n");
945 if (err)
946 goto done;
947
948 err = packet_flush_gently(process->in);
949 if (err)
950 goto done;
951
952 while ((line = packet_read_line(process->out, NULL))) {
953 const char *path;
954 if (skip_prefix(line, "pathname=", &path))
955 string_list_insert(available_paths, xstrdup(path));
956 else
957 ; /* ignore unknown keys */
958 }
959
960 err = subprocess_read_status(process->out, &filter_status);
961 if (err)
962 goto done;
963
964 err = strcmp(filter_status.buf, "success");
965
966 done:
967 sigchain_pop(SIGPIPE);
968
969 if (err)
970 handle_filter_error(&filter_status, entry, 0);
971 strbuf_release(&filter_status);
972 return !err;
973 }
974
975 static struct convert_driver {
976 const char *name;
977 struct convert_driver *next;
978 const char *smudge;
979 const char *clean;
980 const char *process;
981 int required;
982 } *user_convert, **user_convert_tail;
983
984 static int apply_filter(const char *path, const char *src, size_t len,
985 int fd, struct strbuf *dst, struct convert_driver *drv,
986 const unsigned int wanted_capability,
987 const struct checkout_metadata *meta,
988 struct delayed_checkout *dco)
989 {
990 const char *cmd = NULL;
991
992 if (!drv)
993 return 0;
994
995 if (!dst)
996 return 1;
997
998 if ((wanted_capability & CAP_CLEAN) && !drv->process && drv->clean)
999 cmd = drv->clean;
1000 else if ((wanted_capability & CAP_SMUDGE) && !drv->process && drv->smudge)
1001 cmd = drv->smudge;
1002
1003 if (cmd && *cmd)
1004 return apply_single_file_filter(path, src, len, fd, dst, cmd);
1005 else if (drv->process && *drv->process)
1006 return apply_multi_file_filter(path, src, len, fd, dst,
1007 drv->process, wanted_capability, meta, dco);
1008
1009 return 0;
1010 }
1011
1012 static int read_convert_config(const char *var, const char *value, void *cb UNUSED)
1013 {
1014 const char *key, *name;
1015 size_t namelen;
1016 struct convert_driver *drv;
1017
1018 /*
1019 * External conversion drivers are configured using
1020 * "filter.<name>.variable".
1021 */
1022 if (parse_config_key(var, "filter", &name, &namelen, &key) < 0 || !name)
1023 return 0;
1024 for (drv = user_convert; drv; drv = drv->next)
1025 if (!strncmp(drv->name, name, namelen) && !drv->name[namelen])
1026 break;
1027 if (!drv) {
1028 CALLOC_ARRAY(drv, 1);
1029 drv->name = xmemdupz(name, namelen);
1030 *user_convert_tail = drv;
1031 user_convert_tail = &(drv->next);
1032 }
1033
1034 /*
1035 * filter.<name>.smudge and filter.<name>.clean specifies
1036 * the command line:
1037 *
1038 * command-line
1039 *
1040 * The command-line will not be interpolated in any way.
1041 */
1042
1043 if (!strcmp("smudge", key))
1044 return git_config_string(&drv->smudge, var, value);
1045
1046 if (!strcmp("clean", key))
1047 return git_config_string(&drv->clean, var, value);
1048
1049 if (!strcmp("process", key))
1050 return git_config_string(&drv->process, var, value);
1051
1052 if (!strcmp("required", key)) {
1053 drv->required = git_config_bool(var, value);
1054 return 0;
1055 }
1056
1057 return 0;
1058 }
1059
1060 static int count_ident(const char *cp, unsigned long size)
1061 {
1062 /*
1063 * "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$"
1064 */
1065 int cnt = 0;
1066 char ch;
1067
1068 while (size) {
1069 ch = *cp++;
1070 size--;
1071 if (ch != '$')
1072 continue;
1073 if (size < 3)
1074 break;
1075 if (memcmp("Id", cp, 2))
1076 continue;
1077 ch = cp[2];
1078 cp += 3;
1079 size -= 3;
1080 if (ch == '$')
1081 cnt++; /* $Id$ */
1082 if (ch != ':')
1083 continue;
1084
1085 /*
1086 * "$Id: ... "; scan up to the closing dollar sign and discard.
1087 */
1088 while (size) {
1089 ch = *cp++;
1090 size--;
1091 if (ch == '$') {
1092 cnt++;
1093 break;
1094 }
1095 if (ch == '\n')
1096 break;
1097 }
1098 }
1099 return cnt;
1100 }
1101
1102 static int ident_to_git(const char *src, size_t len,
1103 struct strbuf *buf, int ident)
1104 {
1105 char *dst, *dollar;
1106
1107 if (!ident || (src && !count_ident(src, len)))
1108 return 0;
1109
1110 if (!buf)
1111 return 1;
1112
1113 /* only grow if not in place */
1114 if (strbuf_avail(buf) + buf->len < len)
1115 strbuf_grow(buf, len - buf->len);
1116 dst = buf->buf;
1117 for (;;) {
1118 dollar = memchr(src, '$', len);
1119 if (!dollar)
1120 break;
1121 memmove(dst, src, dollar + 1 - src);
1122 dst += dollar + 1 - src;
1123 len -= dollar + 1 - src;
1124 src = dollar + 1;
1125
1126 if (len > 3 && !memcmp(src, "Id:", 3)) {
1127 dollar = memchr(src + 3, '$', len - 3);
1128 if (!dollar)
1129 break;
1130 if (memchr(src + 3, '\n', dollar - src - 3)) {
1131 /* Line break before the next dollar. */
1132 continue;
1133 }
1134
1135 memcpy(dst, "Id$", 3);
1136 dst += 3;
1137 len -= dollar + 1 - src;
1138 src = dollar + 1;
1139 }
1140 }
1141 memmove(dst, src, len);
1142 strbuf_setlen(buf, dst + len - buf->buf);
1143 return 1;
1144 }
1145
1146 static int ident_to_worktree(const char *src, size_t len,
1147 struct strbuf *buf, int ident)
1148 {
1149 struct object_id oid;
1150 char *to_free = NULL, *dollar, *spc;
1151 int cnt;
1152
1153 if (!ident)
1154 return 0;
1155
1156 cnt = count_ident(src, len);
1157 if (!cnt)
1158 return 0;
1159
1160 /* are we "faking" in place editing ? */
1161 if (src == buf->buf)
1162 to_free = strbuf_detach(buf, NULL);
1163 hash_object_file(the_hash_algo, src, len, OBJ_BLOB, &oid);
1164
1165 strbuf_grow(buf, len + cnt * (the_hash_algo->hexsz + 3));
1166 for (;;) {
1167 /* step 1: run to the next '$' */
1168 dollar = memchr(src, '$', len);
1169 if (!dollar)
1170 break;
1171 strbuf_add(buf, src, dollar + 1 - src);
1172 len -= dollar + 1 - src;
1173 src = dollar + 1;
1174
1175 /* step 2: does it looks like a bit like Id:xxx$ or Id$ ? */
1176 if (len < 3 || memcmp("Id", src, 2))
1177 continue;
1178
1179 /* step 3: skip over Id$ or Id:xxxxx$ */
1180 if (src[2] == '$') {
1181 src += 3;
1182 len -= 3;
1183 } else if (src[2] == ':') {
1184 /*
1185 * It's possible that an expanded Id has crept its way into the
1186 * repository, we cope with that by stripping the expansion out.
1187 * This is probably not a good idea, since it will cause changes
1188 * on checkout, which won't go away by stash, but let's keep it
1189 * for git-style ids.
1190 */
1191 dollar = memchr(src + 3, '$', len - 3);
1192 if (!dollar) {
1193 /* incomplete keyword, no more '$', so just quit the loop */
1194 break;
1195 }
1196
1197 if (memchr(src + 3, '\n', dollar - src - 3)) {
1198 /* Line break before the next dollar. */
1199 continue;
1200 }
1201
1202 spc = memchr(src + 4, ' ', dollar - src - 4);
1203 if (spc && spc < dollar-1) {
1204 /* There are spaces in unexpected places.
1205 * This is probably an id from some other
1206 * versioning system. Keep it for now.
1207 */
1208 continue;
1209 }
1210
1211 len -= dollar + 1 - src;
1212 src = dollar + 1;
1213 } else {
1214 /* it wasn't a "Id$" or "Id:xxxx$" */
1215 continue;
1216 }
1217
1218 /* step 4: substitute */
1219 strbuf_addstr(buf, "Id: ");
1220 strbuf_addstr(buf, oid_to_hex(&oid));
1221 strbuf_addstr(buf, " $");
1222 }
1223 strbuf_add(buf, src, len);
1224
1225 free(to_free);
1226 return 1;
1227 }
1228
1229 static const char *git_path_check_encoding(struct attr_check_item *check)
1230 {
1231 const char *value = check->value;
1232
1233 if (ATTR_UNSET(value) || !strlen(value))
1234 return NULL;
1235
1236 if (ATTR_TRUE(value) || ATTR_FALSE(value)) {
1237 die(_("true/false are no valid working-tree-encodings"));
1238 }
1239
1240 /* Don't encode to the default encoding */
1241 if (same_encoding(value, default_encoding))
1242 return NULL;
1243
1244 return value;
1245 }
1246
1247 static enum convert_crlf_action git_path_check_crlf(struct attr_check_item *check)
1248 {
1249 const char *value = check->value;
1250
1251 if (ATTR_TRUE(value))
1252 return CRLF_TEXT;
1253 else if (ATTR_FALSE(value))
1254 return CRLF_BINARY;
1255 else if (ATTR_UNSET(value))
1256 ;
1257 else if (!strcmp(value, "input"))
1258 return CRLF_TEXT_INPUT;
1259 else if (!strcmp(value, "auto"))
1260 return CRLF_AUTO;
1261 return CRLF_UNDEFINED;
1262 }
1263
1264 static enum eol git_path_check_eol(struct attr_check_item *check)
1265 {
1266 const char *value = check->value;
1267
1268 if (ATTR_UNSET(value))
1269 ;
1270 else if (!strcmp(value, "lf"))
1271 return EOL_LF;
1272 else if (!strcmp(value, "crlf"))
1273 return EOL_CRLF;
1274 return EOL_UNSET;
1275 }
1276
1277 static struct convert_driver *git_path_check_convert(struct attr_check_item *check)
1278 {
1279 const char *value = check->value;
1280 struct convert_driver *drv;
1281
1282 if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
1283 return NULL;
1284 for (drv = user_convert; drv; drv = drv->next)
1285 if (!strcmp(value, drv->name))
1286 return drv;
1287 return NULL;
1288 }
1289
1290 static int git_path_check_ident(struct attr_check_item *check)
1291 {
1292 const char *value = check->value;
1293
1294 return !!ATTR_TRUE(value);
1295 }
1296
1297 static struct attr_check *check;
1298
1299 void convert_attrs(struct index_state *istate,
1300 struct conv_attrs *ca, const char *path)
1301 {
1302 struct attr_check_item *ccheck = NULL;
1303
1304 if (!check) {
1305 check = attr_check_initl("crlf", "ident", "filter",
1306 "eol", "text", "working-tree-encoding",
1307 NULL);
1308 user_convert_tail = &user_convert;
1309 git_config(read_convert_config, NULL);
1310 }
1311
1312 git_check_attr(istate, NULL, path, check);
1313 ccheck = check->items;
1314 ca->crlf_action = git_path_check_crlf(ccheck + 4);
1315 if (ca->crlf_action == CRLF_UNDEFINED)
1316 ca->crlf_action = git_path_check_crlf(ccheck + 0);
1317 ca->ident = git_path_check_ident(ccheck + 1);
1318 ca->drv = git_path_check_convert(ccheck + 2);
1319 if (ca->crlf_action != CRLF_BINARY) {
1320 enum eol eol_attr = git_path_check_eol(ccheck + 3);
1321 if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_LF)
1322 ca->crlf_action = CRLF_AUTO_INPUT;
1323 else if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_CRLF)
1324 ca->crlf_action = CRLF_AUTO_CRLF;
1325 else if (eol_attr == EOL_LF)
1326 ca->crlf_action = CRLF_TEXT_INPUT;
1327 else if (eol_attr == EOL_CRLF)
1328 ca->crlf_action = CRLF_TEXT_CRLF;
1329 }
1330 ca->working_tree_encoding = git_path_check_encoding(ccheck + 5);
1331
1332 /* Save attr and make a decision for action */
1333 ca->attr_action = ca->crlf_action;
1334 if (ca->crlf_action == CRLF_TEXT)
1335 ca->crlf_action = text_eol_is_crlf() ? CRLF_TEXT_CRLF : CRLF_TEXT_INPUT;
1336 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_FALSE)
1337 ca->crlf_action = CRLF_BINARY;
1338 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_TRUE)
1339 ca->crlf_action = CRLF_AUTO_CRLF;
1340 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_INPUT)
1341 ca->crlf_action = CRLF_AUTO_INPUT;
1342 }
1343
1344 void reset_parsed_attributes(void)
1345 {
1346 struct convert_driver *drv, *next;
1347
1348 attr_check_free(check);
1349 check = NULL;
1350 reset_merge_attributes();
1351
1352 for (drv = user_convert; drv; drv = next) {
1353 next = drv->next;
1354 free((void *)drv->name);
1355 free(drv);
1356 }
1357 user_convert = NULL;
1358 user_convert_tail = NULL;
1359 }
1360
1361 int would_convert_to_git_filter_fd(struct index_state *istate, const char *path)
1362 {
1363 struct conv_attrs ca;
1364
1365 convert_attrs(istate, &ca, path);
1366 if (!ca.drv)
1367 return 0;
1368
1369 /*
1370 * Apply a filter to an fd only if the filter is required to succeed.
1371 * We must die if the filter fails, because the original data before
1372 * filtering is not available.
1373 */
1374 if (!ca.drv->required)
1375 return 0;
1376
1377 return apply_filter(path, NULL, 0, -1, NULL, ca.drv, CAP_CLEAN, NULL, NULL);
1378 }
1379
1380 const char *get_convert_attr_ascii(struct index_state *istate, const char *path)
1381 {
1382 struct conv_attrs ca;
1383
1384 convert_attrs(istate, &ca, path);
1385 switch (ca.attr_action) {
1386 case CRLF_UNDEFINED:
1387 return "";
1388 case CRLF_BINARY:
1389 return "-text";
1390 case CRLF_TEXT:
1391 return "text";
1392 case CRLF_TEXT_INPUT:
1393 return "text eol=lf";
1394 case CRLF_TEXT_CRLF:
1395 return "text eol=crlf";
1396 case CRLF_AUTO:
1397 return "text=auto";
1398 case CRLF_AUTO_CRLF:
1399 return "text=auto eol=crlf";
1400 case CRLF_AUTO_INPUT:
1401 return "text=auto eol=lf";
1402 }
1403 return "";
1404 }
1405
1406 int convert_to_git(struct index_state *istate,
1407 const char *path, const char *src, size_t len,
1408 struct strbuf *dst, int conv_flags)
1409 {
1410 int ret = 0;
1411 struct conv_attrs ca;
1412
1413 convert_attrs(istate, &ca, path);
1414
1415 ret |= apply_filter(path, src, len, -1, dst, ca.drv, CAP_CLEAN, NULL, NULL);
1416 if (!ret && ca.drv && ca.drv->required)
1417 die(_("%s: clean filter '%s' failed"), path, ca.drv->name);
1418
1419 if (ret && dst) {
1420 src = dst->buf;
1421 len = dst->len;
1422 }
1423
1424 ret |= encode_to_git(path, src, len, dst, ca.working_tree_encoding, conv_flags);
1425 if (ret && dst) {
1426 src = dst->buf;
1427 len = dst->len;
1428 }
1429
1430 if (!(conv_flags & CONV_EOL_KEEP_CRLF)) {
1431 ret |= crlf_to_git(istate, path, src, len, dst, ca.crlf_action, conv_flags);
1432 if (ret && dst) {
1433 src = dst->buf;
1434 len = dst->len;
1435 }
1436 }
1437 return ret | ident_to_git(src, len, dst, ca.ident);
1438 }
1439
1440 void convert_to_git_filter_fd(struct index_state *istate,
1441 const char *path, int fd, struct strbuf *dst,
1442 int conv_flags)
1443 {
1444 struct conv_attrs ca;
1445 convert_attrs(istate, &ca, path);
1446
1447 assert(ca.drv);
1448
1449 if (!apply_filter(path, NULL, 0, fd, dst, ca.drv, CAP_CLEAN, NULL, NULL))
1450 die(_("%s: clean filter '%s' failed"), path, ca.drv->name);
1451
1452 encode_to_git(path, dst->buf, dst->len, dst, ca.working_tree_encoding, conv_flags);
1453 crlf_to_git(istate, path, dst->buf, dst->len, dst, ca.crlf_action, conv_flags);
1454 ident_to_git(dst->buf, dst->len, dst, ca.ident);
1455 }
1456
1457 static int convert_to_working_tree_ca_internal(const struct conv_attrs *ca,
1458 const char *path, const char *src,
1459 size_t len, struct strbuf *dst,
1460 int normalizing,
1461 const struct checkout_metadata *meta,
1462 struct delayed_checkout *dco)
1463 {
1464 int ret = 0, ret_filter = 0;
1465
1466 ret |= ident_to_worktree(src, len, dst, ca->ident);
1467 if (ret) {
1468 src = dst->buf;
1469 len = dst->len;
1470 }
1471 /*
1472 * CRLF conversion can be skipped if normalizing, unless there
1473 * is a smudge or process filter (even if the process filter doesn't
1474 * support smudge). The filters might expect CRLFs.
1475 */
1476 if ((ca->drv && (ca->drv->smudge || ca->drv->process)) || !normalizing) {
1477 ret |= crlf_to_worktree(src, len, dst, ca->crlf_action);
1478 if (ret) {
1479 src = dst->buf;
1480 len = dst->len;
1481 }
1482 }
1483
1484 ret |= encode_to_worktree(path, src, len, dst, ca->working_tree_encoding);
1485 if (ret) {
1486 src = dst->buf;
1487 len = dst->len;
1488 }
1489
1490 ret_filter = apply_filter(
1491 path, src, len, -1, dst, ca->drv, CAP_SMUDGE, meta, dco);
1492 if (!ret_filter && ca->drv && ca->drv->required)
1493 die(_("%s: smudge filter %s failed"), path, ca->drv->name);
1494
1495 return ret | ret_filter;
1496 }
1497
1498 int async_convert_to_working_tree_ca(const struct conv_attrs *ca,
1499 const char *path, const char *src,
1500 size_t len, struct strbuf *dst,
1501 const struct checkout_metadata *meta,
1502 void *dco)
1503 {
1504 return convert_to_working_tree_ca_internal(ca, path, src, len, dst, 0,
1505 meta, dco);
1506 }
1507
1508 int convert_to_working_tree_ca(const struct conv_attrs *ca,
1509 const char *path, const char *src,
1510 size_t len, struct strbuf *dst,
1511 const struct checkout_metadata *meta)
1512 {
1513 return convert_to_working_tree_ca_internal(ca, path, src, len, dst, 0,
1514 meta, NULL);
1515 }
1516
1517 int renormalize_buffer(struct index_state *istate, const char *path,
1518 const char *src, size_t len, struct strbuf *dst)
1519 {
1520 struct conv_attrs ca;
1521 int ret;
1522
1523 convert_attrs(istate, &ca, path);
1524 ret = convert_to_working_tree_ca_internal(&ca, path, src, len, dst, 1,
1525 NULL, NULL);
1526 if (ret) {
1527 src = dst->buf;
1528 len = dst->len;
1529 }
1530 return ret | convert_to_git(istate, path, src, len, dst, CONV_EOL_RENORMALIZE);
1531 }
1532
1533 /*****************************************************************
1534 *
1535 * Streaming conversion support
1536 *
1537 *****************************************************************/
1538
1539 typedef int (*filter_fn)(struct stream_filter *,
1540 const char *input, size_t *isize_p,
1541 char *output, size_t *osize_p);
1542 typedef void (*free_fn)(struct stream_filter *);
1543
1544 struct stream_filter_vtbl {
1545 filter_fn filter;
1546 free_fn free;
1547 };
1548
1549 struct stream_filter {
1550 struct stream_filter_vtbl *vtbl;
1551 };
1552
1553 static int null_filter_fn(struct stream_filter *filter UNUSED,
1554 const char *input, size_t *isize_p,
1555 char *output, size_t *osize_p)
1556 {
1557 size_t count;
1558
1559 if (!input)
1560 return 0; /* we do not keep any states */
1561 count = *isize_p;
1562 if (*osize_p < count)
1563 count = *osize_p;
1564 if (count) {
1565 memmove(output, input, count);
1566 *isize_p -= count;
1567 *osize_p -= count;
1568 }
1569 return 0;
1570 }
1571
1572 static void null_free_fn(struct stream_filter *filter UNUSED)
1573 {
1574 ; /* nothing -- null instances are shared */
1575 }
1576
1577 static struct stream_filter_vtbl null_vtbl = {
1578 .filter = null_filter_fn,
1579 .free = null_free_fn,
1580 };
1581
1582 static struct stream_filter null_filter_singleton = {
1583 .vtbl = &null_vtbl,
1584 };
1585
1586 int is_null_stream_filter(struct stream_filter *filter)
1587 {
1588 return filter == &null_filter_singleton;
1589 }
1590
1591
1592 /*
1593 * LF-to-CRLF filter
1594 */
1595
1596 struct lf_to_crlf_filter {
1597 struct stream_filter filter;
1598 unsigned has_held:1;
1599 char held;
1600 };
1601
1602 static int lf_to_crlf_filter_fn(struct stream_filter *filter,
1603 const char *input, size_t *isize_p,
1604 char *output, size_t *osize_p)
1605 {
1606 size_t count, o = 0;
1607 struct lf_to_crlf_filter *lf_to_crlf = (struct lf_to_crlf_filter *)filter;
1608
1609 /*
1610 * We may be holding onto the CR to see if it is followed by a
1611 * LF, in which case we would need to go to the main loop.
1612 * Otherwise, just emit it to the output stream.
1613 */
1614 if (lf_to_crlf->has_held && (lf_to_crlf->held != '\r' || !input)) {
1615 output[o++] = lf_to_crlf->held;
1616 lf_to_crlf->has_held = 0;
1617 }
1618
1619 /* We are told to drain */
1620 if (!input) {
1621 *osize_p -= o;
1622 return 0;
1623 }
1624
1625 count = *isize_p;
1626 if (count || lf_to_crlf->has_held) {
1627 size_t i;
1628 int was_cr = 0;
1629
1630 if (lf_to_crlf->has_held) {
1631 was_cr = 1;
1632 lf_to_crlf->has_held = 0;
1633 }
1634
1635 for (i = 0; o < *osize_p && i < count; i++) {
1636 char ch = input[i];
1637
1638 if (ch == '\n') {
1639 output[o++] = '\r';
1640 } else if (was_cr) {
1641 /*
1642 * Previous round saw CR and it is not followed
1643 * by a LF; emit the CR before processing the
1644 * current character.
1645 */
1646 output[o++] = '\r';
1647 }
1648
1649 /*
1650 * We may have consumed the last output slot,
1651 * in which case we need to break out of this
1652 * loop; hold the current character before
1653 * returning.
1654 */
1655 if (*osize_p <= o) {
1656 lf_to_crlf->has_held = 1;
1657 lf_to_crlf->held = ch;
1658 continue; /* break but increment i */
1659 }
1660
1661 if (ch == '\r') {
1662 was_cr = 1;
1663 continue;
1664 }
1665
1666 was_cr = 0;
1667 output[o++] = ch;
1668 }
1669
1670 *osize_p -= o;
1671 *isize_p -= i;
1672
1673 if (!lf_to_crlf->has_held && was_cr) {
1674 lf_to_crlf->has_held = 1;
1675 lf_to_crlf->held = '\r';
1676 }
1677 }
1678 return 0;
1679 }
1680
1681 static void lf_to_crlf_free_fn(struct stream_filter *filter)
1682 {
1683 free(filter);
1684 }
1685
1686 static struct stream_filter_vtbl lf_to_crlf_vtbl = {
1687 .filter = lf_to_crlf_filter_fn,
1688 .free = lf_to_crlf_free_fn,
1689 };
1690
1691 static struct stream_filter *lf_to_crlf_filter(void)
1692 {
1693 struct lf_to_crlf_filter *lf_to_crlf = xcalloc(1, sizeof(*lf_to_crlf));
1694
1695 lf_to_crlf->filter.vtbl = &lf_to_crlf_vtbl;
1696 return (struct stream_filter *)lf_to_crlf;
1697 }
1698
1699 /*
1700 * Cascade filter
1701 */
1702 #define FILTER_BUFFER 1024
1703 struct cascade_filter {
1704 struct stream_filter filter;
1705 struct stream_filter *one;
1706 struct stream_filter *two;
1707 char buf[FILTER_BUFFER];
1708 int end, ptr;
1709 };
1710
1711 static int cascade_filter_fn(struct stream_filter *filter,
1712 const char *input, size_t *isize_p,
1713 char *output, size_t *osize_p)
1714 {
1715 struct cascade_filter *cas = (struct cascade_filter *) filter;
1716 size_t filled = 0;
1717 size_t sz = *osize_p;
1718 size_t to_feed, remaining;
1719
1720 /*
1721 * input -- (one) --> buf -- (two) --> output
1722 */
1723 while (filled < sz) {
1724 remaining = sz - filled;
1725
1726 /* do we already have something to feed two with? */
1727 if (cas->ptr < cas->end) {
1728 to_feed = cas->end - cas->ptr;
1729 if (stream_filter(cas->two,
1730 cas->buf + cas->ptr, &to_feed,
1731 output + filled, &remaining))
1732 return -1;
1733 cas->ptr += (cas->end - cas->ptr) - to_feed;
1734 filled = sz - remaining;
1735 continue;
1736 }
1737
1738 /* feed one from upstream and have it emit into our buffer */
1739 to_feed = input ? *isize_p : 0;
1740 if (input && !to_feed)
1741 break;
1742 remaining = sizeof(cas->buf);
1743 if (stream_filter(cas->one,
1744 input, &to_feed,
1745 cas->buf, &remaining))
1746 return -1;
1747 cas->end = sizeof(cas->buf) - remaining;
1748 cas->ptr = 0;
1749 if (input) {
1750 size_t fed = *isize_p - to_feed;
1751 *isize_p -= fed;
1752 input += fed;
1753 }
1754
1755 /* do we know that we drained one completely? */
1756 if (input || cas->end)
1757 continue;
1758
1759 /* tell two to drain; we have nothing more to give it */
1760 to_feed = 0;
1761 remaining = sz - filled;
1762 if (stream_filter(cas->two,
1763 NULL, &to_feed,
1764 output + filled, &remaining))
1765 return -1;
1766 if (remaining == (sz - filled))
1767 break; /* completely drained two */
1768 filled = sz - remaining;
1769 }
1770 *osize_p -= filled;
1771 return 0;
1772 }
1773
1774 static void cascade_free_fn(struct stream_filter *filter)
1775 {
1776 struct cascade_filter *cas = (struct cascade_filter *)filter;
1777 free_stream_filter(cas->one);
1778 free_stream_filter(cas->two);
1779 free(filter);
1780 }
1781
1782 static struct stream_filter_vtbl cascade_vtbl = {
1783 .filter = cascade_filter_fn,
1784 .free = cascade_free_fn,
1785 };
1786
1787 static struct stream_filter *cascade_filter(struct stream_filter *one,
1788 struct stream_filter *two)
1789 {
1790 struct cascade_filter *cascade;
1791
1792 if (!one || is_null_stream_filter(one))
1793 return two;
1794 if (!two || is_null_stream_filter(two))
1795 return one;
1796
1797 cascade = xmalloc(sizeof(*cascade));
1798 cascade->one = one;
1799 cascade->two = two;
1800 cascade->end = cascade->ptr = 0;
1801 cascade->filter.vtbl = &cascade_vtbl;
1802 return (struct stream_filter *)cascade;
1803 }
1804
1805 /*
1806 * ident filter
1807 */
1808 #define IDENT_DRAINING (-1)
1809 #define IDENT_SKIPPING (-2)
1810 struct ident_filter {
1811 struct stream_filter filter;
1812 struct strbuf left;
1813 int state;
1814 char ident[GIT_MAX_HEXSZ + 5]; /* ": x40 $" */
1815 };
1816
1817 static int is_foreign_ident(const char *str)
1818 {
1819 int i;
1820
1821 if (!skip_prefix(str, "$Id: ", &str))
1822 return 0;
1823 for (i = 0; str[i]; i++) {
1824 if (isspace(str[i]) && str[i+1] != '$')
1825 return 1;
1826 }
1827 return 0;
1828 }
1829
1830 static void ident_drain(struct ident_filter *ident, char **output_p, size_t *osize_p)
1831 {
1832 size_t to_drain = ident->left.len;
1833
1834 if (*osize_p < to_drain)
1835 to_drain = *osize_p;
1836 if (to_drain) {
1837 memcpy(*output_p, ident->left.buf, to_drain);
1838 strbuf_remove(&ident->left, 0, to_drain);
1839 *output_p += to_drain;
1840 *osize_p -= to_drain;
1841 }
1842 if (!ident->left.len)
1843 ident->state = 0;
1844 }
1845
1846 static int ident_filter_fn(struct stream_filter *filter,
1847 const char *input, size_t *isize_p,
1848 char *output, size_t *osize_p)
1849 {
1850 struct ident_filter *ident = (struct ident_filter *)filter;
1851 static const char head[] = "$Id";
1852
1853 if (!input) {
1854 /* drain upon eof */
1855 switch (ident->state) {
1856 default:
1857 strbuf_add(&ident->left, head, ident->state);
1858 /* fallthrough */
1859 case IDENT_SKIPPING:
1860 /* fallthrough */
1861 case IDENT_DRAINING:
1862 ident_drain(ident, &output, osize_p);
1863 }
1864 return 0;
1865 }
1866
1867 while (*isize_p || (ident->state == IDENT_DRAINING)) {
1868 int ch;
1869
1870 if (ident->state == IDENT_DRAINING) {
1871 ident_drain(ident, &output, osize_p);
1872 if (!*osize_p)
1873 break;
1874 continue;
1875 }
1876
1877 ch = *(input++);
1878 (*isize_p)--;
1879
1880 if (ident->state == IDENT_SKIPPING) {
1881 /*
1882 * Skipping until '$' or LF, but keeping them
1883 * in case it is a foreign ident.
1884 */
1885 strbuf_addch(&ident->left, ch);
1886 if (ch != '\n' && ch != '$')
1887 continue;
1888 if (ch == '$' && !is_foreign_ident(ident->left.buf)) {
1889 strbuf_setlen(&ident->left, sizeof(head) - 1);
1890 strbuf_addstr(&ident->left, ident->ident);
1891 }
1892 ident->state = IDENT_DRAINING;
1893 continue;
1894 }
1895
1896 if (ident->state < sizeof(head) &&
1897 head[ident->state] == ch) {
1898 ident->state++;
1899 continue;
1900 }
1901
1902 if (ident->state)
1903 strbuf_add(&ident->left, head, ident->state);
1904 if (ident->state == sizeof(head) - 1) {
1905 if (ch != ':' && ch != '$') {
1906 strbuf_addch(&ident->left, ch);
1907 ident->state = 0;
1908 continue;
1909 }
1910
1911 if (ch == ':') {
1912 strbuf_addch(&ident->left, ch);
1913 ident->state = IDENT_SKIPPING;
1914 } else {
1915 strbuf_addstr(&ident->left, ident->ident);
1916 ident->state = IDENT_DRAINING;
1917 }
1918 continue;
1919 }
1920
1921 strbuf_addch(&ident->left, ch);
1922 ident->state = IDENT_DRAINING;
1923 }
1924 return 0;
1925 }
1926
1927 static void ident_free_fn(struct stream_filter *filter)
1928 {
1929 struct ident_filter *ident = (struct ident_filter *)filter;
1930 strbuf_release(&ident->left);
1931 free(filter);
1932 }
1933
1934 static struct stream_filter_vtbl ident_vtbl = {
1935 .filter = ident_filter_fn,
1936 .free = ident_free_fn,
1937 };
1938
1939 static struct stream_filter *ident_filter(const struct object_id *oid)
1940 {
1941 struct ident_filter *ident = xmalloc(sizeof(*ident));
1942
1943 xsnprintf(ident->ident, sizeof(ident->ident),
1944 ": %s $", oid_to_hex(oid));
1945 strbuf_init(&ident->left, 0);
1946 ident->filter.vtbl = &ident_vtbl;
1947 ident->state = 0;
1948 return (struct stream_filter *)ident;
1949 }
1950
1951 /*
1952 * Return an appropriately constructed filter for the given ca, or NULL if
1953 * the contents cannot be filtered without reading the whole thing
1954 * in-core.
1955 *
1956 * Note that you would be crazy to set CRLF, smudge/clean or ident to a
1957 * large binary blob you would want us not to slurp into the memory!
1958 */
1959 struct stream_filter *get_stream_filter_ca(const struct conv_attrs *ca,
1960 const struct object_id *oid)
1961 {
1962 struct stream_filter *filter = NULL;
1963
1964 if (classify_conv_attrs(ca) != CA_CLASS_STREAMABLE)
1965 return NULL;
1966
1967 if (ca->ident)
1968 filter = ident_filter(oid);
1969
1970 if (output_eol(ca->crlf_action) == EOL_CRLF)
1971 filter = cascade_filter(filter, lf_to_crlf_filter());
1972 else
1973 filter = cascade_filter(filter, &null_filter_singleton);
1974
1975 return filter;
1976 }
1977
1978 struct stream_filter *get_stream_filter(struct index_state *istate,
1979 const char *path,
1980 const struct object_id *oid)
1981 {
1982 struct conv_attrs ca;
1983 convert_attrs(istate, &ca, path);
1984 return get_stream_filter_ca(&ca, oid);
1985 }
1986
1987 void free_stream_filter(struct stream_filter *filter)
1988 {
1989 filter->vtbl->free(filter);
1990 }
1991
1992 int stream_filter(struct stream_filter *filter,
1993 const char *input, size_t *isize_p,
1994 char *output, size_t *osize_p)
1995 {
1996 return filter->vtbl->filter(filter, input, isize_p, output, osize_p);
1997 }
1998
1999 void init_checkout_metadata(struct checkout_metadata *meta, const char *refname,
2000 const struct object_id *treeish,
2001 const struct object_id *blob)
2002 {
2003 memset(meta, 0, sizeof(*meta));
2004 if (refname)
2005 meta->refname = refname;
2006 if (treeish)
2007 oidcpy(&meta->treeish, treeish);
2008 if (blob)
2009 oidcpy(&meta->blob, blob);
2010 }
2011
2012 void clone_checkout_metadata(struct checkout_metadata *dst,
2013 const struct checkout_metadata *src,
2014 const struct object_id *blob)
2015 {
2016 memcpy(dst, src, sizeof(*dst));
2017 if (blob)
2018 oidcpy(&dst->blob, blob);
2019 }
2020
2021 enum conv_attrs_classification classify_conv_attrs(const struct conv_attrs *ca)
2022 {
2023 if (ca->drv) {
2024 if (ca->drv->process)
2025 return CA_CLASS_INCORE_PROCESS;
2026 if (ca->drv->smudge || ca->drv->clean)
2027 return CA_CLASS_INCORE_FILTER;
2028 }
2029
2030 if (ca->working_tree_encoding)
2031 return CA_CLASS_INCORE;
2032
2033 if (ca->crlf_action == CRLF_AUTO || ca->crlf_action == CRLF_AUTO_CRLF)
2034 return CA_CLASS_INCORE;
2035
2036 return CA_CLASS_STREAMABLE;
2037 }
git for Debian