?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
sequencer: handle multi-byte comment characters when writing todo list
[git.git] / convert.c
blob35b25eb3cb9212f92dd918d0f247e49ecd620c41
1 #include "git-compat-util.h"
2 #include "advice.h"
3 #include "config.h"
4 #include "convert.h"
5 #include "copy.h"
6 #include "gettext.h"
7 #include "hex.h"
8 #include "object-store-ll.h"
9 #include "attr.h"
10 #include "run-command.h"
11 #include "quote.h"
12 #include "read-cache-ll.h"
13 #include "sigchain.h"
14 #include "pkt-line.h"
15 #include "sub-process.h"
16 #include "trace.h"
17 #include "utf8.h"
18 #include "merge-ll.h"
19
20 /*
21 * convert.c - convert a file when checking it out and checking it in.
22 *
23 * This should use the pathname to decide on whether it wants to do some
24 * more interesting conversions (automatic gzip/unzip, general format
25 * conversions etc etc), but by default it just does automatic CRLF<->LF
26 * translation when the "text" attribute or "auto_crlf" option is set.
27 */
28
29 /* Stat bits: When BIN is set, the txt bits are unset */
30 #define CONVERT_STAT_BITS_TXT_LF 0x1
31 #define CONVERT_STAT_BITS_TXT_CRLF 0x2
32 #define CONVERT_STAT_BITS_BIN 0x4
33
34 struct text_stat {
35 /* NUL, CR, LF and CRLF counts */
36 unsigned nul, lonecr, lonelf, crlf;
37
38 /* These are just approximations! */
39 unsigned printable, nonprintable;
40 };
41
42 static void gather_stats(const char *buf, unsigned long size, struct text_stat *stats)
43 {
44 unsigned long i;
45
46 memset(stats, 0, sizeof(*stats));
47
48 for (i = 0; i < size; i++) {
49 unsigned char c = buf[i];
50 if (c == '\r') {
51 if (i+1 < size && buf[i+1] == '\n') {
52 stats->crlf++;
53 i++;
54 } else
55 stats->lonecr++;
56 continue;
57 }
58 if (c == '\n') {
59 stats->lonelf++;
60 continue;
61 }
62 if (c == 127)
63 /* DEL */
64 stats->nonprintable++;
65 else if (c < 32) {
66 switch (c) {
67 /* BS, HT, ESC and FF */
68 case '\b': case '\t': case '\033': case '\014':
69 stats->printable++;
70 break;
71 case 0:
72 stats->nul++;
73 /* fall through */
74 default:
75 stats->nonprintable++;
76 }
77 }
78 else
79 stats->printable++;
80 }
81
82 /* If file ends with EOF then don't count this EOF as non-printable. */
83 if (size >= 1 && buf[size-1] == '\032')
84 stats->nonprintable--;
85 }
86
87 /*
88 * The same heuristics as diff.c::mmfile_is_binary()
89 * We treat files with bare CR as binary
90 */
91 static int convert_is_binary(const struct text_stat *stats)
92 {
93 if (stats->lonecr)
94 return 1;
95 if (stats->nul)
96 return 1;
97 if ((stats->printable >> 7) < stats->nonprintable)
98 return 1;
99 return 0;
100 }
101
102 static unsigned int gather_convert_stats(const char *data, unsigned long size)
103 {
104 struct text_stat stats;
105 int ret = 0;
106 if (!data || !size)
107 return 0;
108 gather_stats(data, size, &stats);
109 if (convert_is_binary(&stats))
110 ret |= CONVERT_STAT_BITS_BIN;
111 if (stats.crlf)
112 ret |= CONVERT_STAT_BITS_TXT_CRLF;
113 if (stats.lonelf)
114 ret |= CONVERT_STAT_BITS_TXT_LF;
115
116 return ret;
117 }
118
119 static const char *gather_convert_stats_ascii(const char *data, unsigned long size)
120 {
121 unsigned int convert_stats = gather_convert_stats(data, size);
122
123 if (convert_stats & CONVERT_STAT_BITS_BIN)
124 return "-text";
125 switch (convert_stats) {
126 case CONVERT_STAT_BITS_TXT_LF:
127 return "lf";
128 case CONVERT_STAT_BITS_TXT_CRLF:
129 return "crlf";
130 case CONVERT_STAT_BITS_TXT_LF | CONVERT_STAT_BITS_TXT_CRLF:
131 return "mixed";
132 default:
133 return "none";
134 }
135 }
136
137 const char *get_cached_convert_stats_ascii(struct index_state *istate,
138 const char *path)
139 {
140 const char *ret;
141 unsigned long sz;
142 void *data = read_blob_data_from_index(istate, path, &sz);
143 ret = gather_convert_stats_ascii(data, sz);
144 free(data);
145 return ret;
146 }
147
148 const char *get_wt_convert_stats_ascii(const char *path)
149 {
150 const char *ret = "";
151 struct strbuf sb = STRBUF_INIT;
152 if (strbuf_read_file(&sb, path, 0) >= 0)
153 ret = gather_convert_stats_ascii(sb.buf, sb.len);
154 strbuf_release(&sb);
155 return ret;
156 }
157
158 static int text_eol_is_crlf(void)
159 {
160 if (auto_crlf == AUTO_CRLF_TRUE)
161 return 1;
162 else if (auto_crlf == AUTO_CRLF_INPUT)
163 return 0;
164 if (core_eol == EOL_CRLF)
165 return 1;
166 if (core_eol == EOL_UNSET && EOL_NATIVE == EOL_CRLF)
167 return 1;
168 return 0;
169 }
170
171 static enum eol output_eol(enum convert_crlf_action crlf_action)
172 {
173 switch (crlf_action) {
174 case CRLF_BINARY:
175 return EOL_UNSET;
176 case CRLF_TEXT_CRLF:
177 return EOL_CRLF;
178 case CRLF_TEXT_INPUT:
179 return EOL_LF;
180 case CRLF_UNDEFINED:
181 case CRLF_AUTO_CRLF:
182 return EOL_CRLF;
183 case CRLF_AUTO_INPUT:
184 return EOL_LF;
185 case CRLF_TEXT:
186 case CRLF_AUTO:
187 /* fall through */
188 return text_eol_is_crlf() ? EOL_CRLF : EOL_LF;
189 }
190 warning(_("illegal crlf_action %d"), (int)crlf_action);
191 return core_eol;
192 }
193
194 static void check_global_conv_flags_eol(const char *path,
195 struct text_stat *old_stats, struct text_stat *new_stats,
196 int conv_flags)
197 {
198 if (old_stats->crlf && !new_stats->crlf ) {
199 /*
200 * CRLFs would not be restored by checkout
201 */
202 if (conv_flags & CONV_EOL_RNDTRP_DIE)
203 die(_("CRLF would be replaced by LF in %s"), path);
204 else if (conv_flags & CONV_EOL_RNDTRP_WARN)
205 warning(_("in the working copy of '%s', CRLF will be"
206 " replaced by LF the next time Git touches"
207 " it"), path);
208 } else if (old_stats->lonelf && !new_stats->lonelf ) {
209 /*
210 * CRLFs would be added by checkout
211 */
212 if (conv_flags & CONV_EOL_RNDTRP_DIE)
213 die(_("LF would be replaced by CRLF in %s"), path);
214 else if (conv_flags & CONV_EOL_RNDTRP_WARN)
215 warning(_("in the working copy of '%s', LF will be"
216 " replaced by CRLF the next time Git touches"
217 " it"), path);
218 }
219 }
220
221 static int has_crlf_in_index(struct index_state *istate, const char *path)
222 {
223 unsigned long sz;
224 void *data;
225 const char *crp;
226 int has_crlf = 0;
227
228 data = read_blob_data_from_index(istate, path, &sz);
229 if (!data)
230 return 0;
231
232 crp = memchr(data, '\r', sz);
233 if (crp) {
234 unsigned int ret_stats;
235 ret_stats = gather_convert_stats(data, sz);
236 if (!(ret_stats & CONVERT_STAT_BITS_BIN) &&
237 (ret_stats & CONVERT_STAT_BITS_TXT_CRLF))
238 has_crlf = 1;
239 }
240 free(data);
241 return has_crlf;
242 }
243
244 static int will_convert_lf_to_crlf(struct text_stat *stats,
245 enum convert_crlf_action crlf_action)
246 {
247 if (output_eol(crlf_action) != EOL_CRLF)
248 return 0;
249 /* No "naked" LF? Nothing to convert, regardless. */
250 if (!stats->lonelf)
251 return 0;
252
253 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
254 /* If we have any CR or CRLF line endings, we do not touch it */
255 /* This is the new safer autocrlf-handling */
256 if (stats->lonecr || stats->crlf)
257 return 0;
258
259 if (convert_is_binary(stats))
260 return 0;
261 }
262 return 1;
263
264 }
265
266 static int validate_encoding(const char *path, const char *enc,
267 const char *data, size_t len, int die_on_error)
268 {
269 const char *stripped;
270
271 /* We only check for UTF here as UTF?? can be an alias for UTF-?? */
272 if (skip_iprefix(enc, "UTF", &stripped)) {
273 skip_prefix(stripped, "-", &stripped);
274
275 /*
276 * Check for detectable errors in UTF encodings
277 */
278 if (has_prohibited_utf_bom(enc, data, len)) {
279 const char *error_msg = _(
280 "BOM is prohibited in '%s' if encoded as %s");
281 /*
282 * This advice is shown for UTF-??BE and UTF-??LE encodings.
283 * We cut off the last two characters of the encoding name
284 * to generate the encoding name suitable for BOMs.
285 */
286 const char *advise_msg = _(
287 "The file '%s' contains a byte order "
288 "mark (BOM). Please use UTF-%.*s as "
289 "working-tree-encoding.");
290 int stripped_len = strlen(stripped) - strlen("BE");
291 advise(advise_msg, path, stripped_len, stripped);
292 if (die_on_error)
293 die(error_msg, path, enc);
294 else {
295 return error(error_msg, path, enc);
296 }
297
298 } else if (is_missing_required_utf_bom(enc, data, len)) {
299 const char *error_msg = _(
300 "BOM is required in '%s' if encoded as %s");
301 const char *advise_msg = _(
302 "The file '%s' is missing a byte order "
303 "mark (BOM). Please use UTF-%sBE or UTF-%sLE "
304 "(depending on the byte order) as "
305 "working-tree-encoding.");
306 advise(advise_msg, path, stripped, stripped);
307 if (die_on_error)
308 die(error_msg, path, enc);
309 else {
310 return error(error_msg, path, enc);
311 }
312 }
313
314 }
315 return 0;
316 }
317
318 static void trace_encoding(const char *context, const char *path,
319 const char *encoding, const char *buf, size_t len)
320 {
321 static struct trace_key coe = TRACE_KEY_INIT(WORKING_TREE_ENCODING);
322 struct strbuf trace = STRBUF_INIT;
323 int i;
324
325 strbuf_addf(&trace, "%s (%s, considered %s):\n", context, path, encoding);
326 for (i = 0; i < len && buf; ++i) {
327 strbuf_addf(
328 &trace, "| \033[2m%2i:\033[0m %2x \033[2m%c\033[0m%c",
329 i,
330 (unsigned char) buf[i],
331 (buf[i] > 32 && buf[i] < 127 ? buf[i] : ' '),
332 ((i+1) % 8 && (i+1) < len ? ' ' : '\n')
333 );
334 }
335 strbuf_addchars(&trace, '\n', 1);
336
337 trace_strbuf(&coe, &trace);
338 strbuf_release(&trace);
339 }
340
341 static int check_roundtrip(const char *enc_name)
342 {
343 /*
344 * check_roundtrip_encoding contains a string of comma and/or
345 * space separated encodings (eg. "UTF-16, ASCII, CP1125").
346 * Search for the given encoding in that string.
347 */
348 const char *found = strcasestr(check_roundtrip_encoding, enc_name);
349 const char *next;
350 int len;
351 if (!found)
352 return 0;
353 next = found + strlen(enc_name);
354 len = strlen(check_roundtrip_encoding);
355 return (found && (
356 /*
357 * check that the found encoding is at the
358 * beginning of check_roundtrip_encoding or
359 * that it is prefixed with a space or comma
360 */
361 found == check_roundtrip_encoding || (
362 (isspace(found[-1]) || found[-1] == ',')
363 )
364 ) && (
365 /*
366 * check that the found encoding is at the
367 * end of check_roundtrip_encoding or
368 * that it is suffixed with a space or comma
369 */
370 next == check_roundtrip_encoding + len || (
371 next < check_roundtrip_encoding + len &&
372 (isspace(next[0]) || next[0] == ',')
373 )
374 ));
375 }
376
377 static const char *default_encoding = "UTF-8";
378
379 static int encode_to_git(const char *path, const char *src, size_t src_len,
380 struct strbuf *buf, const char *enc, int conv_flags)
381 {
382 char *dst;
383 size_t dst_len;
384 int die_on_error = conv_flags & CONV_WRITE_OBJECT;
385
386 /*
387 * No encoding is specified or there is nothing to encode.
388 * Tell the caller that the content was not modified.
389 */
390 if (!enc || (src && !src_len))
391 return 0;
392
393 /*
394 * Looks like we got called from "would_convert_to_git()".
395 * This means Git wants to know if it would encode (= modify!)
396 * the content. Let's answer with "yes", since an encoding was
397 * specified.
398 */
399 if (!buf && !src)
400 return 1;
401
402 if (validate_encoding(path, enc, src, src_len, die_on_error))
403 return 0;
404
405 trace_encoding("source", path, enc, src, src_len);
406 dst = reencode_string_len(src, src_len, default_encoding, enc,
407 &dst_len);
408 if (!dst) {
409 /*
410 * We could add the blob "as-is" to Git. However, on checkout
411 * we would try to re-encode to the original encoding. This
412 * would fail and we would leave the user with a messed-up
413 * working tree. Let's try to avoid this by screaming loud.
414 */
415 const char* msg = _("failed to encode '%s' from %s to %s");
416 if (die_on_error)
417 die(msg, path, enc, default_encoding);
418 else {
419 error(msg, path, enc, default_encoding);
420 return 0;
421 }
422 }
423 trace_encoding("destination", path, default_encoding, dst, dst_len);
424
425 /*
426 * UTF supports lossless conversion round tripping [1] and conversions
427 * between UTF and other encodings are mostly round trip safe as
428 * Unicode aims to be a superset of all other character encodings.
429 * However, certain encodings (e.g. SHIFT-JIS) are known to have round
430 * trip issues [2]. Check the round trip conversion for all encodings
431 * listed in core.checkRoundtripEncoding.
432 *
433 * The round trip check is only performed if content is written to Git.
434 * This ensures that no information is lost during conversion to/from
435 * the internal UTF-8 representation.
436 *
437 * Please note, the code below is not tested because I was not able to
438 * generate a faulty round trip without an iconv error. Iconv errors
439 * are already caught above.
440 *
441 * [1] http://unicode.org/faq/utf_bom.html#gen2
442 * [2] https://support.microsoft.com/en-us/help/170559/prb-conversion-problem-between-shift-jis-and-unicode
443 */
444 if (die_on_error && check_roundtrip(enc)) {
445 char *re_src;
446 size_t re_src_len;
447
448 re_src = reencode_string_len(dst, dst_len,
449 enc, default_encoding,
450 &re_src_len);
451
452 trace_printf("Checking roundtrip encoding for %s...\n", enc);
453 trace_encoding("reencoded source", path, enc,
454 re_src, re_src_len);
455
456 if (!re_src || src_len != re_src_len ||
457 memcmp(src, re_src, src_len)) {
458 const char* msg = _("encoding '%s' from %s to %s and "
459 "back is not the same");
460 die(msg, path, enc, default_encoding);
461 }
462
463 free(re_src);
464 }
465
466 strbuf_attach(buf, dst, dst_len, dst_len + 1);
467 return 1;
468 }
469
470 static int encode_to_worktree(const char *path, const char *src, size_t src_len,
471 struct strbuf *buf, const char *enc)
472 {
473 char *dst;
474 size_t dst_len;
475
476 /*
477 * No encoding is specified or there is nothing to encode.
478 * Tell the caller that the content was not modified.
479 */
480 if (!enc || (src && !src_len))
481 return 0;
482
483 dst = reencode_string_len(src, src_len, enc, default_encoding,
484 &dst_len);
485 if (!dst) {
486 error(_("failed to encode '%s' from %s to %s"),
487 path, default_encoding, enc);
488 return 0;
489 }
490
491 strbuf_attach(buf, dst, dst_len, dst_len + 1);
492 return 1;
493 }
494
495 static int crlf_to_git(struct index_state *istate,
496 const char *path, const char *src, size_t len,
497 struct strbuf *buf,
498 enum convert_crlf_action crlf_action, int conv_flags)
499 {
500 struct text_stat stats;
501 char *dst;
502 int convert_crlf_into_lf;
503
504 if (crlf_action == CRLF_BINARY ||
505 (src && !len))
506 return 0;
507
508 /*
509 * If we are doing a dry-run and have no source buffer, there is
510 * nothing to analyze; we must assume we would convert.
511 */
512 if (!buf && !src)
513 return 1;
514
515 gather_stats(src, len, &stats);
516 /* Optimization: No CRLF? Nothing to convert, regardless. */
517 convert_crlf_into_lf = !!stats.crlf;
518
519 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
520 if (convert_is_binary(&stats))
521 return 0;
522 /*
523 * If the file in the index has any CR in it, do not
524 * convert. This is the new safer autocrlf handling,
525 * unless we want to renormalize in a merge or
526 * cherry-pick.
527 */
528 if ((!(conv_flags & CONV_EOL_RENORMALIZE)) &&
529 has_crlf_in_index(istate, path))
530 convert_crlf_into_lf = 0;
531 }
532 if (((conv_flags & CONV_EOL_RNDTRP_WARN) ||
533 ((conv_flags & CONV_EOL_RNDTRP_DIE) && len))) {
534 struct text_stat new_stats;
535 memcpy(&new_stats, &stats, sizeof(new_stats));
536 /* simulate "git add" */
537 if (convert_crlf_into_lf) {
538 new_stats.lonelf += new_stats.crlf;
539 new_stats.crlf = 0;
540 }
541 /* simulate "git checkout" */
542 if (will_convert_lf_to_crlf(&new_stats, crlf_action)) {
543 new_stats.crlf += new_stats.lonelf;
544 new_stats.lonelf = 0;
545 }
546 check_global_conv_flags_eol(path, &stats, &new_stats, conv_flags);
547 }
548 if (!convert_crlf_into_lf)
549 return 0;
550
551 /*
552 * At this point all of our source analysis is done, and we are sure we
553 * would convert. If we are in dry-run mode, we can give an answer.
554 */
555 if (!buf)
556 return 1;
557
558 /* only grow if not in place */
559 if (strbuf_avail(buf) + buf->len < len)
560 strbuf_grow(buf, len - buf->len);
561 dst = buf->buf;
562 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
563 /*
564 * If we guessed, we already know we rejected a file with
565 * lone CR, and we can strip a CR without looking at what
566 * follow it.
567 */
568 do {
569 unsigned char c = *src++;
570 if (c != '\r')
571 *dst++ = c;
572 } while (--len);
573 } else {
574 do {
575 unsigned char c = *src++;
576 if (! (c == '\r' && (1 < len && *src == '\n')))
577 *dst++ = c;
578 } while (--len);
579 }
580 strbuf_setlen(buf, dst - buf->buf);
581 return 1;
582 }
583
584 static int crlf_to_worktree(const char *src, size_t len, struct strbuf *buf,
585 enum convert_crlf_action crlf_action)
586 {
587 char *to_free = NULL;
588 struct text_stat stats;
589
590 if (!len || output_eol(crlf_action) != EOL_CRLF)
591 return 0;
592
593 gather_stats(src, len, &stats);
594 if (!will_convert_lf_to_crlf(&stats, crlf_action))
595 return 0;
596
597 /* are we "faking" in place editing ? */
598 if (src == buf->buf)
599 to_free = strbuf_detach(buf, NULL);
600
601 strbuf_grow(buf, len + stats.lonelf);
602 for (;;) {
603 const char *nl = memchr(src, '\n', len);
604 if (!nl)
605 break;
606 if (nl > src && nl[-1] == '\r') {
607 strbuf_add(buf, src, nl + 1 - src);
608 } else {
609 strbuf_add(buf, src, nl - src);
610 strbuf_addstr(buf, "\r\n");
611 }
612 len -= nl + 1 - src;
613 src = nl + 1;
614 }
615 strbuf_add(buf, src, len);
616
617 free(to_free);
618 return 1;
619 }
620
621 struct filter_params {
622 const char *src;
623 size_t size;
624 int fd;
625 const char *cmd;
626 const char *path;
627 };
628
629 static int filter_buffer_or_fd(int in UNUSED, int out, void *data)
630 {
631 /*
632 * Spawn cmd and feed the buffer contents through its stdin.
633 */
634 struct child_process child_process = CHILD_PROCESS_INIT;
635 struct filter_params *params = (struct filter_params *)data;
636 const char *format = params->cmd;
637 int write_err, status;
638
639 /* apply % substitution to cmd */
640 struct strbuf cmd = STRBUF_INIT;
641
642 /* expand all %f with the quoted path; quote to preserve space, etc. */
643 while (strbuf_expand_step(&cmd, &format)) {
644 if (skip_prefix(format, "%", &format))
645 strbuf_addch(&cmd, '%');
646 else if (skip_prefix(format, "f", &format))
647 sq_quote_buf(&cmd, params->path);
648 else
649 strbuf_addch(&cmd, '%');
650 }
651
652 strvec_push(&child_process.args, cmd.buf);
653 child_process.use_shell = 1;
654 child_process.in = -1;
655 child_process.out = out;
656
657 if (start_command(&child_process)) {
658 strbuf_release(&cmd);
659 return error(_("cannot fork to run external filter '%s'"),
660 params->cmd);
661 }
662
663 sigchain_push(SIGPIPE, SIG_IGN);
664
665 if (params->src) {
666 write_err = (write_in_full(child_process.in,
667 params->src, params->size) < 0);
668 if (errno == EPIPE)
669 write_err = 0;
670 } else {
671 write_err = copy_fd(params->fd, child_process.in);
672 if (write_err == COPY_WRITE_ERROR && errno == EPIPE)
673 write_err = 0;
674 }
675
676 if (close(child_process.in))
677 write_err = 1;
678 if (write_err)
679 error(_("cannot feed the input to external filter '%s'"),
680 params->cmd);
681
682 sigchain_pop(SIGPIPE);
683
684 status = finish_command(&child_process);
685 if (status)
686 error(_("external filter '%s' failed %d"), params->cmd, status);
687
688 strbuf_release(&cmd);
689 return (write_err || status);
690 }
691
692 static int apply_single_file_filter(const char *path, const char *src, size_t len, int fd,
693 struct strbuf *dst, const char *cmd)
694 {
695 /*
696 * Create a pipeline to have the command filter the buffer's
697 * contents.
698 *
699 * (child --> cmd) --> us
700 */
701 int err = 0;
702 struct strbuf nbuf = STRBUF_INIT;
703 struct async async;
704 struct filter_params params;
705
706 memset(&async, 0, sizeof(async));
707 async.proc = filter_buffer_or_fd;
708 async.data = &params;
709 async.out = -1;
710 params.src = src;
711 params.size = len;
712 params.fd = fd;
713 params.cmd = cmd;
714 params.path = path;
715
716 fflush(NULL);
717 if (start_async(&async))
718 return 0; /* error was already reported */
719
720 if (strbuf_read(&nbuf, async.out, 0) < 0) {
721 err = error(_("read from external filter '%s' failed"), cmd);
722 }
723 if (close(async.out)) {
724 err = error(_("read from external filter '%s' failed"), cmd);
725 }
726 if (finish_async(&async)) {
727 err = error(_("external filter '%s' failed"), cmd);
728 }
729
730 if (!err) {
731 strbuf_swap(dst, &nbuf);
732 }
733 strbuf_release(&nbuf);
734 return !err;
735 }
736
737 #define CAP_CLEAN (1u<<0)
738 #define CAP_SMUDGE (1u<<1)
739 #define CAP_DELAY (1u<<2)
740
741 struct cmd2process {
742 struct subprocess_entry subprocess; /* must be the first member! */
743 unsigned int supported_capabilities;
744 };
745
746 static int subprocess_map_initialized;
747 static struct hashmap subprocess_map;
748
749 static int start_multi_file_filter_fn(struct subprocess_entry *subprocess)
750 {
751 static int versions[] = {2, 0};
752 static struct subprocess_capability capabilities[] = {
753 { "clean", CAP_CLEAN },
754 { "smudge", CAP_SMUDGE },
755 { "delay", CAP_DELAY },
756 { NULL, 0 }
757 };
758 struct cmd2process *entry = (struct cmd2process *)subprocess;
759 return subprocess_handshake(subprocess, "git-filter", versions, NULL,
760 capabilities,
761 &entry->supported_capabilities);
762 }
763
764 static void handle_filter_error(const struct strbuf *filter_status,
765 struct cmd2process *entry,
766 const unsigned int wanted_capability)
767 {
768 if (!strcmp(filter_status->buf, "error"))
769 ; /* The filter signaled a problem with the file. */
770 else if (!strcmp(filter_status->buf, "abort") && wanted_capability) {
771 /*
772 * The filter signaled a permanent problem. Don't try to filter
773 * files with the same command for the lifetime of the current
774 * Git process.
775 */
776 entry->supported_capabilities &= ~wanted_capability;
777 } else {
778 /*
779 * Something went wrong with the protocol filter.
780 * Force shutdown and restart if another blob requires filtering.
781 */
782 error(_("external filter '%s' failed"), entry->subprocess.cmd);
783 subprocess_stop(&subprocess_map, &entry->subprocess);
784 free(entry);
785 }
786 }
787
788 static int apply_multi_file_filter(const char *path, const char *src, size_t len,
789 int fd, struct strbuf *dst, const char *cmd,
790 const unsigned int wanted_capability,
791 const struct checkout_metadata *meta,
792 struct delayed_checkout *dco)
793 {
794 int err;
795 int can_delay = 0;
796 struct cmd2process *entry;
797 struct child_process *process;
798 struct strbuf nbuf = STRBUF_INIT;
799 struct strbuf filter_status = STRBUF_INIT;
800 const char *filter_type;
801
802 if (!subprocess_map_initialized) {
803 subprocess_map_initialized = 1;
804 hashmap_init(&subprocess_map, cmd2process_cmp, NULL, 0);
805 entry = NULL;
806 } else {
807 entry = (struct cmd2process *)subprocess_find_entry(&subprocess_map, cmd);
808 }
809
810 fflush(NULL);
811
812 if (!entry) {
813 entry = xmalloc(sizeof(*entry));
814 entry->supported_capabilities = 0;
815
816 if (subprocess_start(&subprocess_map, &entry->subprocess, cmd, start_multi_file_filter_fn)) {
817 free(entry);
818 return 0;
819 }
820 }
821 process = &entry->subprocess.process;
822
823 if (!(entry->supported_capabilities & wanted_capability))
824 return 0;
825
826 if (wanted_capability & CAP_CLEAN)
827 filter_type = "clean";
828 else if (wanted_capability & CAP_SMUDGE)
829 filter_type = "smudge";
830 else
831 die(_("unexpected filter type"));
832
833 sigchain_push(SIGPIPE, SIG_IGN);
834
835 assert(strlen(filter_type) < LARGE_PACKET_DATA_MAX - strlen("command=\n"));
836 err = packet_write_fmt_gently(process->in, "command=%s\n", filter_type);
837 if (err)
838 goto done;
839
840 err = strlen(path) > LARGE_PACKET_DATA_MAX - strlen("pathname=\n");
841 if (err) {
842 error(_("path name too long for external filter"));
843 goto done;
844 }
845
846 err = packet_write_fmt_gently(process->in, "pathname=%s\n", path);
847 if (err)
848 goto done;
849
850 if (meta && meta->refname) {
851 err = packet_write_fmt_gently(process->in, "ref=%s\n", meta->refname);
852 if (err)
853 goto done;
854 }
855
856 if (meta && !is_null_oid(&meta->treeish)) {
857 err = packet_write_fmt_gently(process->in, "treeish=%s\n", oid_to_hex(&meta->treeish));
858 if (err)
859 goto done;
860 }
861
862 if (meta && !is_null_oid(&meta->blob)) {
863 err = packet_write_fmt_gently(process->in, "blob=%s\n", oid_to_hex(&meta->blob));
864 if (err)
865 goto done;
866 }
867
868 if ((entry->supported_capabilities & CAP_DELAY) &&
869 dco && dco->state == CE_CAN_DELAY) {
870 can_delay = 1;
871 err = packet_write_fmt_gently(process->in, "can-delay=1\n");
872 if (err)
873 goto done;
874 }
875
876 err = packet_flush_gently(process->in);
877 if (err)
878 goto done;
879
880 if (fd >= 0)
881 err = write_packetized_from_fd_no_flush(fd, process->in);
882 else
883 err = write_packetized_from_buf_no_flush(src, len, process->in);
884 if (err)
885 goto done;
886
887 err = packet_flush_gently(process->in);
888 if (err)
889 goto done;
890
891 err = subprocess_read_status(process->out, &filter_status);
892 if (err)
893 goto done;
894
895 if (can_delay && !strcmp(filter_status.buf, "delayed")) {
896 string_list_insert(&dco->filters, cmd);
897 string_list_insert(&dco->paths, path);
898 } else {
899 /* The filter got the blob and wants to send us a response. */
900 err = strcmp(filter_status.buf, "success");
901 if (err)
902 goto done;
903
904 err = read_packetized_to_strbuf(process->out, &nbuf,
905 PACKET_READ_GENTLE_ON_EOF) < 0;
906 if (err)
907 goto done;
908
909 err = subprocess_read_status(process->out, &filter_status);
910 if (err)
911 goto done;
912
913 err = strcmp(filter_status.buf, "success");
914 }
915
916 done:
917 sigchain_pop(SIGPIPE);
918
919 if (err)
920 handle_filter_error(&filter_status, entry, wanted_capability);
921 else
922 strbuf_swap(dst, &nbuf);
923 strbuf_release(&nbuf);
924 strbuf_release(&filter_status);
925 return !err;
926 }
927
928
929 int async_query_available_blobs(const char *cmd, struct string_list *available_paths)
930 {
931 int err;
932 char *line;
933 struct cmd2process *entry;
934 struct child_process *process;
935 struct strbuf filter_status = STRBUF_INIT;
936
937 assert(subprocess_map_initialized);
938 entry = (struct cmd2process *)subprocess_find_entry(&subprocess_map, cmd);
939 if (!entry) {
940 error(_("external filter '%s' is not available anymore although "
941 "not all paths have been filtered"), cmd);
942 return 0;
943 }
944 process = &entry->subprocess.process;
945 sigchain_push(SIGPIPE, SIG_IGN);
946
947 err = packet_write_fmt_gently(
948 process->in, "command=list_available_blobs\n");
949 if (err)
950 goto done;
951
952 err = packet_flush_gently(process->in);
953 if (err)
954 goto done;
955
956 while ((line = packet_read_line(process->out, NULL))) {
957 const char *path;
958 if (skip_prefix(line, "pathname=", &path))
959 string_list_insert(available_paths, xstrdup(path));
960 else
961 ; /* ignore unknown keys */
962 }
963
964 err = subprocess_read_status(process->out, &filter_status);
965 if (err)
966 goto done;
967
968 err = strcmp(filter_status.buf, "success");
969
970 done:
971 sigchain_pop(SIGPIPE);
972
973 if (err)
974 handle_filter_error(&filter_status, entry, 0);
975 strbuf_release(&filter_status);
976 return !err;
977 }
978
979 static struct convert_driver {
980 const char *name;
981 struct convert_driver *next;
982 const char *smudge;
983 const char *clean;
984 const char *process;
985 int required;
986 } *user_convert, **user_convert_tail;
987
988 static int apply_filter(const char *path, const char *src, size_t len,
989 int fd, struct strbuf *dst, struct convert_driver *drv,
990 const unsigned int wanted_capability,
991 const struct checkout_metadata *meta,
992 struct delayed_checkout *dco)
993 {
994 const char *cmd = NULL;
995
996 if (!drv)
997 return 0;
998
999 if (!dst)
1000 return 1;
1001
1002 if ((wanted_capability & CAP_CLEAN) && !drv->process && drv->clean)
1003 cmd = drv->clean;
1004 else if ((wanted_capability & CAP_SMUDGE) && !drv->process && drv->smudge)
1005 cmd = drv->smudge;
1006
1007 if (cmd && *cmd)
1008 return apply_single_file_filter(path, src, len, fd, dst, cmd);
1009 else if (drv->process && *drv->process)
1010 return apply_multi_file_filter(path, src, len, fd, dst,
1011 drv->process, wanted_capability, meta, dco);
1012
1013 return 0;
1014 }
1015
1016 static int read_convert_config(const char *var, const char *value,
1017 const struct config_context *ctx UNUSED,
1018 void *cb UNUSED)
1019 {
1020 const char *key, *name;
1021 size_t namelen;
1022 struct convert_driver *drv;
1023
1024 /*
1025 * External conversion drivers are configured using
1026 * "filter.<name>.variable".
1027 */
1028 if (parse_config_key(var, "filter", &name, &namelen, &key) < 0 || !name)
1029 return 0;
1030 for (drv = user_convert; drv; drv = drv->next)
1031 if (!xstrncmpz(drv->name, name, namelen))
1032 break;
1033 if (!drv) {
1034 CALLOC_ARRAY(drv, 1);
1035 drv->name = xmemdupz(name, namelen);
1036 *user_convert_tail = drv;
1037 user_convert_tail = &(drv->next);
1038 }
1039
1040 /*
1041 * filter.<name>.smudge and filter.<name>.clean specifies
1042 * the command line:
1043 *
1044 * command-line
1045 *
1046 * The command-line will not be interpolated in any way.
1047 */
1048
1049 if (!strcmp("smudge", key))
1050 return git_config_string(&drv->smudge, var, value);
1051
1052 if (!strcmp("clean", key))
1053 return git_config_string(&drv->clean, var, value);
1054
1055 if (!strcmp("process", key))
1056 return git_config_string(&drv->process, var, value);
1057
1058 if (!strcmp("required", key)) {
1059 drv->required = git_config_bool(var, value);
1060 return 0;
1061 }
1062
1063 return 0;
1064 }
1065
1066 static int count_ident(const char *cp, unsigned long size)
1067 {
1068 /*
1069 * "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$"
1070 */
1071 int cnt = 0;
1072 char ch;
1073
1074 while (size) {
1075 ch = *cp++;
1076 size--;
1077 if (ch != '$')
1078 continue;
1079 if (size < 3)
1080 break;
1081 if (memcmp("Id", cp, 2))
1082 continue;
1083 ch = cp[2];
1084 cp += 3;
1085 size -= 3;
1086 if (ch == '$')
1087 cnt++; /* $Id$ */
1088 if (ch != ':')
1089 continue;
1090
1091 /*
1092 * "$Id: ... "; scan up to the closing dollar sign and discard.
1093 */
1094 while (size) {
1095 ch = *cp++;
1096 size--;
1097 if (ch == '$') {
1098 cnt++;
1099 break;
1100 }
1101 if (ch == '\n')
1102 break;
1103 }
1104 }
1105 return cnt;
1106 }
1107
1108 static int ident_to_git(const char *src, size_t len,
1109 struct strbuf *buf, int ident)
1110 {
1111 char *dst, *dollar;
1112
1113 if (!ident || (src && !count_ident(src, len)))
1114 return 0;
1115
1116 if (!buf)
1117 return 1;
1118
1119 /* only grow if not in place */
1120 if (strbuf_avail(buf) + buf->len < len)
1121 strbuf_grow(buf, len - buf->len);
1122 dst = buf->buf;
1123 for (;;) {
1124 dollar = memchr(src, '$', len);
1125 if (!dollar)
1126 break;
1127 memmove(dst, src, dollar + 1 - src);
1128 dst += dollar + 1 - src;
1129 len -= dollar + 1 - src;
1130 src = dollar + 1;
1131
1132 if (len > 3 && !memcmp(src, "Id:", 3)) {
1133 dollar = memchr(src + 3, '$', len - 3);
1134 if (!dollar)
1135 break;
1136 if (memchr(src + 3, '\n', dollar - src - 3)) {
1137 /* Line break before the next dollar. */
1138 continue;
1139 }
1140
1141 memcpy(dst, "Id$", 3);
1142 dst += 3;
1143 len -= dollar + 1 - src;
1144 src = dollar + 1;
1145 }
1146 }
1147 memmove(dst, src, len);
1148 strbuf_setlen(buf, dst + len - buf->buf);
1149 return 1;
1150 }
1151
1152 static int ident_to_worktree(const char *src, size_t len,
1153 struct strbuf *buf, int ident)
1154 {
1155 struct object_id oid;
1156 char *to_free = NULL, *dollar, *spc;
1157 int cnt;
1158
1159 if (!ident)
1160 return 0;
1161
1162 cnt = count_ident(src, len);
1163 if (!cnt)
1164 return 0;
1165
1166 /* are we "faking" in place editing ? */
1167 if (src == buf->buf)
1168 to_free = strbuf_detach(buf, NULL);
1169 hash_object_file(the_hash_algo, src, len, OBJ_BLOB, &oid);
1170
1171 strbuf_grow(buf, len + cnt * (the_hash_algo->hexsz + 3));
1172 for (;;) {
1173 /* step 1: run to the next '$' */
1174 dollar = memchr(src, '$', len);
1175 if (!dollar)
1176 break;
1177 strbuf_add(buf, src, dollar + 1 - src);
1178 len -= dollar + 1 - src;
1179 src = dollar + 1;
1180
1181 /* step 2: does it looks like a bit like Id:xxx$ or Id$ ? */
1182 if (len < 3 || memcmp("Id", src, 2))
1183 continue;
1184
1185 /* step 3: skip over Id$ or Id:xxxxx$ */
1186 if (src[2] == '$') {
1187 src += 3;
1188 len -= 3;
1189 } else if (src[2] == ':') {
1190 /*
1191 * It's possible that an expanded Id has crept its way into the
1192 * repository, we cope with that by stripping the expansion out.
1193 * This is probably not a good idea, since it will cause changes
1194 * on checkout, which won't go away by stash, but let's keep it
1195 * for git-style ids.
1196 */
1197 dollar = memchr(src + 3, '$', len - 3);
1198 if (!dollar) {
1199 /* incomplete keyword, no more '$', so just quit the loop */
1200 break;
1201 }
1202
1203 if (memchr(src + 3, '\n', dollar - src - 3)) {
1204 /* Line break before the next dollar. */
1205 continue;
1206 }
1207
1208 spc = memchr(src + 4, ' ', dollar - src - 4);
1209 if (spc && spc < dollar-1) {
1210 /* There are spaces in unexpected places.
1211 * This is probably an id from some other
1212 * versioning system. Keep it for now.
1213 */
1214 continue;
1215 }
1216
1217 len -= dollar + 1 - src;
1218 src = dollar + 1;
1219 } else {
1220 /* it wasn't a "Id$" or "Id:xxxx$" */
1221 continue;
1222 }
1223
1224 /* step 4: substitute */
1225 strbuf_addstr(buf, "Id: ");
1226 strbuf_addstr(buf, oid_to_hex(&oid));
1227 strbuf_addstr(buf, " $");
1228 }
1229 strbuf_add(buf, src, len);
1230
1231 free(to_free);
1232 return 1;
1233 }
1234
1235 static const char *git_path_check_encoding(struct attr_check_item *check)
1236 {
1237 const char *value = check->value;
1238
1239 if (ATTR_UNSET(value) || !strlen(value))
1240 return NULL;
1241
1242 if (ATTR_TRUE(value) || ATTR_FALSE(value)) {
1243 die(_("true/false are no valid working-tree-encodings"));
1244 }
1245
1246 /* Don't encode to the default encoding */
1247 if (same_encoding(value, default_encoding))
1248 return NULL;
1249
1250 return value;
1251 }
1252
1253 static enum convert_crlf_action git_path_check_crlf(struct attr_check_item *check)
1254 {
1255 const char *value = check->value;
1256
1257 if (ATTR_TRUE(value))
1258 return CRLF_TEXT;
1259 else if (ATTR_FALSE(value))
1260 return CRLF_BINARY;
1261 else if (ATTR_UNSET(value))
1262 ;
1263 else if (!strcmp(value, "input"))
1264 return CRLF_TEXT_INPUT;
1265 else if (!strcmp(value, "auto"))
1266 return CRLF_AUTO;
1267 return CRLF_UNDEFINED;
1268 }
1269
1270 static enum eol git_path_check_eol(struct attr_check_item *check)
1271 {
1272 const char *value = check->value;
1273
1274 if (ATTR_UNSET(value))
1275 ;
1276 else if (!strcmp(value, "lf"))
1277 return EOL_LF;
1278 else if (!strcmp(value, "crlf"))
1279 return EOL_CRLF;
1280 return EOL_UNSET;
1281 }
1282
1283 static struct convert_driver *git_path_check_convert(struct attr_check_item *check)
1284 {
1285 const char *value = check->value;
1286 struct convert_driver *drv;
1287
1288 if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
1289 return NULL;
1290 for (drv = user_convert; drv; drv = drv->next)
1291 if (!strcmp(value, drv->name))
1292 return drv;
1293 return NULL;
1294 }
1295
1296 static int git_path_check_ident(struct attr_check_item *check)
1297 {
1298 const char *value = check->value;
1299
1300 return !!ATTR_TRUE(value);
1301 }
1302
1303 static struct attr_check *check;
1304
1305 void convert_attrs(struct index_state *istate,
1306 struct conv_attrs *ca, const char *path)
1307 {
1308 struct attr_check_item *ccheck = NULL;
1309
1310 if (!check) {
1311 check = attr_check_initl("crlf", "ident", "filter",
1312 "eol", "text", "working-tree-encoding",
1313 NULL);
1314 user_convert_tail = &user_convert;
1315 git_config(read_convert_config, NULL);
1316 }
1317
1318 git_check_attr(istate, path, check);
1319 ccheck = check->items;
1320 ca->crlf_action = git_path_check_crlf(ccheck + 4);
1321 if (ca->crlf_action == CRLF_UNDEFINED)
1322 ca->crlf_action = git_path_check_crlf(ccheck + 0);
1323 ca->ident = git_path_check_ident(ccheck + 1);
1324 ca->drv = git_path_check_convert(ccheck + 2);
1325 if (ca->crlf_action != CRLF_BINARY) {
1326 enum eol eol_attr = git_path_check_eol(ccheck + 3);
1327 if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_LF)
1328 ca->crlf_action = CRLF_AUTO_INPUT;
1329 else if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_CRLF)
1330 ca->crlf_action = CRLF_AUTO_CRLF;
1331 else if (eol_attr == EOL_LF)
1332 ca->crlf_action = CRLF_TEXT_INPUT;
1333 else if (eol_attr == EOL_CRLF)
1334 ca->crlf_action = CRLF_TEXT_CRLF;
1335 }
1336 ca->working_tree_encoding = git_path_check_encoding(ccheck + 5);
1337
1338 /* Save attr and make a decision for action */
1339 ca->attr_action = ca->crlf_action;
1340 if (ca->crlf_action == CRLF_TEXT)
1341 ca->crlf_action = text_eol_is_crlf() ? CRLF_TEXT_CRLF : CRLF_TEXT_INPUT;
1342 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_FALSE)
1343 ca->crlf_action = CRLF_BINARY;
1344 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_TRUE)
1345 ca->crlf_action = CRLF_AUTO_CRLF;
1346 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_INPUT)
1347 ca->crlf_action = CRLF_AUTO_INPUT;
1348 }
1349
1350 void reset_parsed_attributes(void)
1351 {
1352 struct convert_driver *drv, *next;
1353
1354 attr_check_free(check);
1355 check = NULL;
1356 reset_merge_attributes();
1357
1358 for (drv = user_convert; drv; drv = next) {
1359 next = drv->next;
1360 free((void *)drv->name);
1361 free(drv);
1362 }
1363 user_convert = NULL;
1364 user_convert_tail = NULL;
1365 }
1366
1367 int would_convert_to_git_filter_fd(struct index_state *istate, const char *path)
1368 {
1369 struct conv_attrs ca;
1370
1371 convert_attrs(istate, &ca, path);
1372 if (!ca.drv)
1373 return 0;
1374
1375 /*
1376 * Apply a filter to an fd only if the filter is required to succeed.
1377 * We must die if the filter fails, because the original data before
1378 * filtering is not available.
1379 */
1380 if (!ca.drv->required)
1381 return 0;
1382
1383 return apply_filter(path, NULL, 0, -1, NULL, ca.drv, CAP_CLEAN, NULL, NULL);
1384 }
1385
1386 const char *get_convert_attr_ascii(struct index_state *istate, const char *path)
1387 {
1388 struct conv_attrs ca;
1389
1390 convert_attrs(istate, &ca, path);
1391 switch (ca.attr_action) {
1392 case CRLF_UNDEFINED:
1393 return "";
1394 case CRLF_BINARY:
1395 return "-text";
1396 case CRLF_TEXT:
1397 return "text";
1398 case CRLF_TEXT_INPUT:
1399 return "text eol=lf";
1400 case CRLF_TEXT_CRLF:
1401 return "text eol=crlf";
1402 case CRLF_AUTO:
1403 return "text=auto";
1404 case CRLF_AUTO_CRLF:
1405 return "text=auto eol=crlf";
1406 case CRLF_AUTO_INPUT:
1407 return "text=auto eol=lf";
1408 }
1409 return "";
1410 }
1411
1412 int convert_to_git(struct index_state *istate,
1413 const char *path, const char *src, size_t len,
1414 struct strbuf *dst, int conv_flags)
1415 {
1416 int ret = 0;
1417 struct conv_attrs ca;
1418
1419 convert_attrs(istate, &ca, path);
1420
1421 ret |= apply_filter(path, src, len, -1, dst, ca.drv, CAP_CLEAN, NULL, NULL);
1422 if (!ret && ca.drv && ca.drv->required)
1423 die(_("%s: clean filter '%s' failed"), path, ca.drv->name);
1424
1425 if (ret && dst) {
1426 src = dst->buf;
1427 len = dst->len;
1428 }
1429
1430 ret |= encode_to_git(path, src, len, dst, ca.working_tree_encoding, conv_flags);
1431 if (ret && dst) {
1432 src = dst->buf;
1433 len = dst->len;
1434 }
1435
1436 if (!(conv_flags & CONV_EOL_KEEP_CRLF)) {
1437 ret |= crlf_to_git(istate, path, src, len, dst, ca.crlf_action, conv_flags);
1438 if (ret && dst) {
1439 src = dst->buf;
1440 len = dst->len;
1441 }
1442 }
1443 return ret | ident_to_git(src, len, dst, ca.ident);
1444 }
1445
1446 void convert_to_git_filter_fd(struct index_state *istate,
1447 const char *path, int fd, struct strbuf *dst,
1448 int conv_flags)
1449 {
1450 struct conv_attrs ca;
1451 convert_attrs(istate, &ca, path);
1452
1453 assert(ca.drv);
1454
1455 if (!apply_filter(path, NULL, 0, fd, dst, ca.drv, CAP_CLEAN, NULL, NULL))
1456 die(_("%s: clean filter '%s' failed"), path, ca.drv->name);
1457
1458 encode_to_git(path, dst->buf, dst->len, dst, ca.working_tree_encoding, conv_flags);
1459 crlf_to_git(istate, path, dst->buf, dst->len, dst, ca.crlf_action, conv_flags);
1460 ident_to_git(dst->buf, dst->len, dst, ca.ident);
1461 }
1462
1463 static int convert_to_working_tree_ca_internal(const struct conv_attrs *ca,
1464 const char *path, const char *src,
1465 size_t len, struct strbuf *dst,
1466 int normalizing,
1467 const struct checkout_metadata *meta,
1468 struct delayed_checkout *dco)
1469 {
1470 int ret = 0, ret_filter = 0;
1471
1472 ret |= ident_to_worktree(src, len, dst, ca->ident);
1473 if (ret) {
1474 src = dst->buf;
1475 len = dst->len;
1476 }
1477 /*
1478 * CRLF conversion can be skipped if normalizing, unless there
1479 * is a smudge or process filter (even if the process filter doesn't
1480 * support smudge). The filters might expect CRLFs.
1481 */
1482 if ((ca->drv && (ca->drv->smudge || ca->drv->process)) || !normalizing) {
1483 ret |= crlf_to_worktree(src, len, dst, ca->crlf_action);
1484 if (ret) {
1485 src = dst->buf;
1486 len = dst->len;
1487 }
1488 }
1489
1490 ret |= encode_to_worktree(path, src, len, dst, ca->working_tree_encoding);
1491 if (ret) {
1492 src = dst->buf;
1493 len = dst->len;
1494 }
1495
1496 ret_filter = apply_filter(
1497 path, src, len, -1, dst, ca->drv, CAP_SMUDGE, meta, dco);
1498 if (!ret_filter && ca->drv && ca->drv->required)
1499 die(_("%s: smudge filter %s failed"), path, ca->drv->name);
1500
1501 return ret | ret_filter;
1502 }
1503
1504 int async_convert_to_working_tree_ca(const struct conv_attrs *ca,
1505 const char *path, const char *src,
1506 size_t len, struct strbuf *dst,
1507 const struct checkout_metadata *meta,
1508 void *dco)
1509 {
1510 return convert_to_working_tree_ca_internal(ca, path, src, len, dst, 0,
1511 meta, dco);
1512 }
1513
1514 int convert_to_working_tree_ca(const struct conv_attrs *ca,
1515 const char *path, const char *src,
1516 size_t len, struct strbuf *dst,
1517 const struct checkout_metadata *meta)
1518 {
1519 return convert_to_working_tree_ca_internal(ca, path, src, len, dst, 0,
1520 meta, NULL);
1521 }
1522
1523 int renormalize_buffer(struct index_state *istate, const char *path,
1524 const char *src, size_t len, struct strbuf *dst)
1525 {
1526 struct conv_attrs ca;
1527 int ret;
1528
1529 convert_attrs(istate, &ca, path);
1530 ret = convert_to_working_tree_ca_internal(&ca, path, src, len, dst, 1,
1531 NULL, NULL);
1532 if (ret) {
1533 src = dst->buf;
1534 len = dst->len;
1535 }
1536 return ret | convert_to_git(istate, path, src, len, dst, CONV_EOL_RENORMALIZE);
1537 }
1538
1539 /*****************************************************************
1540 *
1541 * Streaming conversion support
1542 *
1543 *****************************************************************/
1544
1545 typedef int (*filter_fn)(struct stream_filter *,
1546 const char *input, size_t *isize_p,
1547 char *output, size_t *osize_p);
1548 typedef void (*free_fn)(struct stream_filter *);
1549
1550 struct stream_filter_vtbl {
1551 filter_fn filter;
1552 free_fn free;
1553 };
1554
1555 struct stream_filter {
1556 struct stream_filter_vtbl *vtbl;
1557 };
1558
1559 static int null_filter_fn(struct stream_filter *filter UNUSED,
1560 const char *input, size_t *isize_p,
1561 char *output, size_t *osize_p)
1562 {
1563 size_t count;
1564
1565 if (!input)
1566 return 0; /* we do not keep any states */
1567 count = *isize_p;
1568 if (*osize_p < count)
1569 count = *osize_p;
1570 if (count) {
1571 memmove(output, input, count);
1572 *isize_p -= count;
1573 *osize_p -= count;
1574 }
1575 return 0;
1576 }
1577
1578 static void null_free_fn(struct stream_filter *filter UNUSED)
1579 {
1580 ; /* nothing -- null instances are shared */
1581 }
1582
1583 static struct stream_filter_vtbl null_vtbl = {
1584 .filter = null_filter_fn,
1585 .free = null_free_fn,
1586 };
1587
1588 static struct stream_filter null_filter_singleton = {
1589 .vtbl = &null_vtbl,
1590 };
1591
1592 int is_null_stream_filter(struct stream_filter *filter)
1593 {
1594 return filter == &null_filter_singleton;
1595 }
1596
1597
1598 /*
1599 * LF-to-CRLF filter
1600 */
1601
1602 struct lf_to_crlf_filter {
1603 struct stream_filter filter;
1604 unsigned has_held:1;
1605 char held;
1606 };
1607
1608 static int lf_to_crlf_filter_fn(struct stream_filter *filter,
1609 const char *input, size_t *isize_p,
1610 char *output, size_t *osize_p)
1611 {
1612 size_t count, o = 0;
1613 struct lf_to_crlf_filter *lf_to_crlf = (struct lf_to_crlf_filter *)filter;
1614
1615 /*
1616 * We may be holding onto the CR to see if it is followed by a
1617 * LF, in which case we would need to go to the main loop.
1618 * Otherwise, just emit it to the output stream.
1619 */
1620 if (lf_to_crlf->has_held && (lf_to_crlf->held != '\r' || !input)) {
1621 output[o++] = lf_to_crlf->held;
1622 lf_to_crlf->has_held = 0;
1623 }
1624
1625 /* We are told to drain */
1626 if (!input) {
1627 *osize_p -= o;
1628 return 0;
1629 }
1630
1631 count = *isize_p;
1632 if (count || lf_to_crlf->has_held) {
1633 size_t i;
1634 int was_cr = 0;
1635
1636 if (lf_to_crlf->has_held) {
1637 was_cr = 1;
1638 lf_to_crlf->has_held = 0;
1639 }
1640
1641 for (i = 0; o < *osize_p && i < count; i++) {
1642 char ch = input[i];
1643
1644 if (ch == '\n') {
1645 output[o++] = '\r';
1646 } else if (was_cr) {
1647 /*
1648 * Previous round saw CR and it is not followed
1649 * by a LF; emit the CR before processing the
1650 * current character.
1651 */
1652 output[o++] = '\r';
1653 }
1654
1655 /*
1656 * We may have consumed the last output slot,
1657 * in which case we need to break out of this
1658 * loop; hold the current character before
1659 * returning.
1660 */
1661 if (*osize_p <= o) {
1662 lf_to_crlf->has_held = 1;
1663 lf_to_crlf->held = ch;
1664 continue; /* break but increment i */
1665 }
1666
1667 if (ch == '\r') {
1668 was_cr = 1;
1669 continue;
1670 }
1671
1672 was_cr = 0;
1673 output[o++] = ch;
1674 }
1675
1676 *osize_p -= o;
1677 *isize_p -= i;
1678
1679 if (!lf_to_crlf->has_held && was_cr) {
1680 lf_to_crlf->has_held = 1;
1681 lf_to_crlf->held = '\r';
1682 }
1683 }
1684 return 0;
1685 }
1686
1687 static void lf_to_crlf_free_fn(struct stream_filter *filter)
1688 {
1689 free(filter);
1690 }
1691
1692 static struct stream_filter_vtbl lf_to_crlf_vtbl = {
1693 .filter = lf_to_crlf_filter_fn,
1694 .free = lf_to_crlf_free_fn,
1695 };
1696
1697 static struct stream_filter *lf_to_crlf_filter(void)
1698 {
1699 struct lf_to_crlf_filter *lf_to_crlf = xcalloc(1, sizeof(*lf_to_crlf));
1700
1701 lf_to_crlf->filter.vtbl = &lf_to_crlf_vtbl;
1702 return (struct stream_filter *)lf_to_crlf;
1703 }
1704
1705 /*
1706 * Cascade filter
1707 */
1708 #define FILTER_BUFFER 1024
1709 struct cascade_filter {
1710 struct stream_filter filter;
1711 struct stream_filter *one;
1712 struct stream_filter *two;
1713 char buf[FILTER_BUFFER];
1714 int end, ptr;
1715 };
1716
1717 static int cascade_filter_fn(struct stream_filter *filter,
1718 const char *input, size_t *isize_p,
1719 char *output, size_t *osize_p)
1720 {
1721 struct cascade_filter *cas = (struct cascade_filter *) filter;
1722 size_t filled = 0;
1723 size_t sz = *osize_p;
1724 size_t to_feed, remaining;
1725
1726 /*
1727 * input -- (one) --> buf -- (two) --> output
1728 */
1729 while (filled < sz) {
1730 remaining = sz - filled;
1731
1732 /* do we already have something to feed two with? */
1733 if (cas->ptr < cas->end) {
1734 to_feed = cas->end - cas->ptr;
1735 if (stream_filter(cas->two,
1736 cas->buf + cas->ptr, &to_feed,
1737 output + filled, &remaining))
1738 return -1;
1739 cas->ptr += (cas->end - cas->ptr) - to_feed;
1740 filled = sz - remaining;
1741 continue;
1742 }
1743
1744 /* feed one from upstream and have it emit into our buffer */
1745 to_feed = input ? *isize_p : 0;
1746 if (input && !to_feed)
1747 break;
1748 remaining = sizeof(cas->buf);
1749 if (stream_filter(cas->one,
1750 input, &to_feed,
1751 cas->buf, &remaining))
1752 return -1;
1753 cas->end = sizeof(cas->buf) - remaining;
1754 cas->ptr = 0;
1755 if (input) {
1756 size_t fed = *isize_p - to_feed;
1757 *isize_p -= fed;
1758 input += fed;
1759 }
1760
1761 /* do we know that we drained one completely? */
1762 if (input || cas->end)
1763 continue;
1764
1765 /* tell two to drain; we have nothing more to give it */
1766 to_feed = 0;
1767 remaining = sz - filled;
1768 if (stream_filter(cas->two,
1769 NULL, &to_feed,
1770 output + filled, &remaining))
1771 return -1;
1772 if (remaining == (sz - filled))
1773 break; /* completely drained two */
1774 filled = sz - remaining;
1775 }
1776 *osize_p -= filled;
1777 return 0;
1778 }
1779
1780 static void cascade_free_fn(struct stream_filter *filter)
1781 {
1782 struct cascade_filter *cas = (struct cascade_filter *)filter;
1783 free_stream_filter(cas->one);
1784 free_stream_filter(cas->two);
1785 free(filter);
1786 }
1787
1788 static struct stream_filter_vtbl cascade_vtbl = {
1789 .filter = cascade_filter_fn,
1790 .free = cascade_free_fn,
1791 };
1792
1793 static struct stream_filter *cascade_filter(struct stream_filter *one,
1794 struct stream_filter *two)
1795 {
1796 struct cascade_filter *cascade;
1797
1798 if (!one || is_null_stream_filter(one))
1799 return two;
1800 if (!two || is_null_stream_filter(two))
1801 return one;
1802
1803 cascade = xmalloc(sizeof(*cascade));
1804 cascade->one = one;
1805 cascade->two = two;
1806 cascade->end = cascade->ptr = 0;
1807 cascade->filter.vtbl = &cascade_vtbl;
1808 return (struct stream_filter *)cascade;
1809 }
1810
1811 /*
1812 * ident filter
1813 */
1814 #define IDENT_DRAINING (-1)
1815 #define IDENT_SKIPPING (-2)
1816 struct ident_filter {
1817 struct stream_filter filter;
1818 struct strbuf left;
1819 int state;
1820 char ident[GIT_MAX_HEXSZ + 5]; /* ": x40 $" */
1821 };
1822
1823 static int is_foreign_ident(const char *str)
1824 {
1825 int i;
1826
1827 if (!skip_prefix(str, "$Id: ", &str))
1828 return 0;
1829 for (i = 0; str[i]; i++) {
1830 if (isspace(str[i]) && str[i+1] != '$')
1831 return 1;
1832 }
1833 return 0;
1834 }
1835
1836 static void ident_drain(struct ident_filter *ident, char **output_p, size_t *osize_p)
1837 {
1838 size_t to_drain = ident->left.len;
1839
1840 if (*osize_p < to_drain)
1841 to_drain = *osize_p;
1842 if (to_drain) {
1843 memcpy(*output_p, ident->left.buf, to_drain);
1844 strbuf_remove(&ident->left, 0, to_drain);
1845 *output_p += to_drain;
1846 *osize_p -= to_drain;
1847 }
1848 if (!ident->left.len)
1849 ident->state = 0;
1850 }
1851
1852 static int ident_filter_fn(struct stream_filter *filter,
1853 const char *input, size_t *isize_p,
1854 char *output, size_t *osize_p)
1855 {
1856 struct ident_filter *ident = (struct ident_filter *)filter;
1857 static const char head[] = "$Id";
1858
1859 if (!input) {
1860 /* drain upon eof */
1861 switch (ident->state) {
1862 default:
1863 strbuf_add(&ident->left, head, ident->state);
1864 /* fallthrough */
1865 case IDENT_SKIPPING:
1866 /* fallthrough */
1867 case IDENT_DRAINING:
1868 ident_drain(ident, &output, osize_p);
1869 }
1870 return 0;
1871 }
1872
1873 while (*isize_p || (ident->state == IDENT_DRAINING)) {
1874 int ch;
1875
1876 if (ident->state == IDENT_DRAINING) {
1877 ident_drain(ident, &output, osize_p);
1878 if (!*osize_p)
1879 break;
1880 continue;
1881 }
1882
1883 ch = *(input++);
1884 (*isize_p)--;
1885
1886 if (ident->state == IDENT_SKIPPING) {
1887 /*
1888 * Skipping until '$' or LF, but keeping them
1889 * in case it is a foreign ident.
1890 */
1891 strbuf_addch(&ident->left, ch);
1892 if (ch != '\n' && ch != '$')
1893 continue;
1894 if (ch == '$' && !is_foreign_ident(ident->left.buf)) {
1895 strbuf_setlen(&ident->left, sizeof(head) - 1);
1896 strbuf_addstr(&ident->left, ident->ident);
1897 }
1898 ident->state = IDENT_DRAINING;
1899 continue;
1900 }
1901
1902 if (ident->state < sizeof(head) &&
1903 head[ident->state] == ch) {
1904 ident->state++;
1905 continue;
1906 }
1907
1908 if (ident->state)
1909 strbuf_add(&ident->left, head, ident->state);
1910 if (ident->state == sizeof(head) - 1) {
1911 if (ch != ':' && ch != '$') {
1912 strbuf_addch(&ident->left, ch);
1913 ident->state = 0;
1914 continue;
1915 }
1916
1917 if (ch == ':') {
1918 strbuf_addch(&ident->left, ch);
1919 ident->state = IDENT_SKIPPING;
1920 } else {
1921 strbuf_addstr(&ident->left, ident->ident);
1922 ident->state = IDENT_DRAINING;
1923 }
1924 continue;
1925 }
1926
1927 strbuf_addch(&ident->left, ch);
1928 ident->state = IDENT_DRAINING;
1929 }
1930 return 0;
1931 }
1932
1933 static void ident_free_fn(struct stream_filter *filter)
1934 {
1935 struct ident_filter *ident = (struct ident_filter *)filter;
1936 strbuf_release(&ident->left);
1937 free(filter);
1938 }
1939
1940 static struct stream_filter_vtbl ident_vtbl = {
1941 .filter = ident_filter_fn,
1942 .free = ident_free_fn,
1943 };
1944
1945 static struct stream_filter *ident_filter(const struct object_id *oid)
1946 {
1947 struct ident_filter *ident = xmalloc(sizeof(*ident));
1948
1949 xsnprintf(ident->ident, sizeof(ident->ident),
1950 ": %s $", oid_to_hex(oid));
1951 strbuf_init(&ident->left, 0);
1952 ident->filter.vtbl = &ident_vtbl;
1953 ident->state = 0;
1954 return (struct stream_filter *)ident;
1955 }
1956
1957 /*
1958 * Return an appropriately constructed filter for the given ca, or NULL if
1959 * the contents cannot be filtered without reading the whole thing
1960 * in-core.
1961 *
1962 * Note that you would be crazy to set CRLF, smudge/clean or ident to a
1963 * large binary blob you would want us not to slurp into the memory!
1964 */
1965 struct stream_filter *get_stream_filter_ca(const struct conv_attrs *ca,
1966 const struct object_id *oid)
1967 {
1968 struct stream_filter *filter = NULL;
1969
1970 if (classify_conv_attrs(ca) != CA_CLASS_STREAMABLE)
1971 return NULL;
1972
1973 if (ca->ident)
1974 filter = ident_filter(oid);
1975
1976 if (output_eol(ca->crlf_action) == EOL_CRLF)
1977 filter = cascade_filter(filter, lf_to_crlf_filter());
1978 else
1979 filter = cascade_filter(filter, &null_filter_singleton);
1980
1981 return filter;
1982 }
1983
1984 struct stream_filter *get_stream_filter(struct index_state *istate,
1985 const char *path,
1986 const struct object_id *oid)
1987 {
1988 struct conv_attrs ca;
1989 convert_attrs(istate, &ca, path);
1990 return get_stream_filter_ca(&ca, oid);
1991 }
1992
1993 void free_stream_filter(struct stream_filter *filter)
1994 {
1995 filter->vtbl->free(filter);
1996 }
1997
1998 int stream_filter(struct stream_filter *filter,
1999 const char *input, size_t *isize_p,
2000 char *output, size_t *osize_p)
2001 {
2002 return filter->vtbl->filter(filter, input, isize_p, output, osize_p);
2003 }
2004
2005 void init_checkout_metadata(struct checkout_metadata *meta, const char *refname,
2006 const struct object_id *treeish,
2007 const struct object_id *blob)
2008 {
2009 memset(meta, 0, sizeof(*meta));
2010 if (refname)
2011 meta->refname = refname;
2012 if (treeish)
2013 oidcpy(&meta->treeish, treeish);
2014 if (blob)
2015 oidcpy(&meta->blob, blob);
2016 }
2017
2018 void clone_checkout_metadata(struct checkout_metadata *dst,
2019 const struct checkout_metadata *src,
2020 const struct object_id *blob)
2021 {
2022 memcpy(dst, src, sizeof(*dst));
2023 if (blob)
2024 oidcpy(&dst->blob, blob);
2025 }
2026
2027 enum conv_attrs_classification classify_conv_attrs(const struct conv_attrs *ca)
2028 {
2029 if (ca->drv) {
2030 if (ca->drv->process)
2031 return CA_CLASS_INCORE_PROCESS;
2032 if (ca->drv->smudge || ca->drv->clean)
2033 return CA_CLASS_INCORE_FILTER;
2034 }
2035
2036 if (ca->working_tree_encoding)
2037 return CA_CLASS_INCORE;
2038
2039 if (ca->crlf_action == CRLF_AUTO || ca->crlf_action == CRLF_AUTO_CRLF)
2040 return CA_CLASS_INCORE;
2041
2042 return CA_CLASS_STREAMABLE;
2043 }
The Best VCS Around