convert: separate generic structures and variables from the filter specific ones
[git/git.git] / convert.c
1 #include "cache.h"
2 #include "attr.h"
3 #include "run-command.h"
4 #include "quote.h"
5 #include "sigchain.h"
6 #include "pkt-line.h"
7
8 /*
9 * convert.c - convert a file when checking it out and checking it in.
10 *
11 * This should use the pathname to decide on whether it wants to do some
12 * more interesting conversions (automatic gzip/unzip, general format
13 * conversions etc etc), but by default it just does automatic CRLF<->LF
14 * translation when the "text" attribute or "auto_crlf" option is set.
15 */
16
17 /* Stat bits: When BIN is set, the txt bits are unset */
18 #define CONVERT_STAT_BITS_TXT_LF 0x1
19 #define CONVERT_STAT_BITS_TXT_CRLF 0x2
20 #define CONVERT_STAT_BITS_BIN 0x4
21
22 enum crlf_action {
23 CRLF_UNDEFINED,
24 CRLF_BINARY,
25 CRLF_TEXT,
26 CRLF_TEXT_INPUT,
27 CRLF_TEXT_CRLF,
28 CRLF_AUTO,
29 CRLF_AUTO_INPUT,
30 CRLF_AUTO_CRLF
31 };
32
33 struct text_stat {
34 /* NUL, CR, LF and CRLF counts */
35 unsigned nul, lonecr, lonelf, crlf;
36
37 /* These are just approximations! */
38 unsigned printable, nonprintable;
39 };
40
41 static void gather_stats(const char *buf, unsigned long size, struct text_stat *stats)
42 {
43 unsigned long i;
44
45 memset(stats, 0, sizeof(*stats));
46
47 for (i = 0; i < size; i++) {
48 unsigned char c = buf[i];
49 if (c == '\r') {
50 if (i+1 < size && buf[i+1] == '\n') {
51 stats->crlf++;
52 i++;
53 } else
54 stats->lonecr++;
55 continue;
56 }
57 if (c == '\n') {
58 stats->lonelf++;
59 continue;
60 }
61 if (c == 127)
62 /* DEL */
63 stats->nonprintable++;
64 else if (c < 32) {
65 switch (c) {
66 /* BS, HT, ESC and FF */
67 case '\b': case '\t': case '\033': case '\014':
68 stats->printable++;
69 break;
70 case 0:
71 stats->nul++;
72 /* fall through */
73 default:
74 stats->nonprintable++;
75 }
76 }
77 else
78 stats->printable++;
79 }
80
81 /* If file ends with EOF then don't count this EOF as non-printable. */
82 if (size >= 1 && buf[size-1] == '\032')
83 stats->nonprintable--;
84 }
85
86 /*
87 * The same heuristics as diff.c::mmfile_is_binary()
88 * We treat files with bare CR as binary
89 */
90 static int convert_is_binary(unsigned long size, const struct text_stat *stats)
91 {
92 if (stats->lonecr)
93 return 1;
94 if (stats->nul)
95 return 1;
96 if ((stats->printable >> 7) < stats->nonprintable)
97 return 1;
98 return 0;
99 }
100
101 static unsigned int gather_convert_stats(const char *data, unsigned long size)
102 {
103 struct text_stat stats;
104 int ret = 0;
105 if (!data || !size)
106 return 0;
107 gather_stats(data, size, &stats);
108 if (convert_is_binary(size, &stats))
109 ret |= CONVERT_STAT_BITS_BIN;
110 if (stats.crlf)
111 ret |= CONVERT_STAT_BITS_TXT_CRLF;
112 if (stats.lonelf)
113 ret |= CONVERT_STAT_BITS_TXT_LF;
114
115 return ret;
116 }
117
118 static const char *gather_convert_stats_ascii(const char *data, unsigned long size)
119 {
120 unsigned int convert_stats = gather_convert_stats(data, size);
121
122 if (convert_stats & CONVERT_STAT_BITS_BIN)
123 return "-text";
124 switch (convert_stats) {
125 case CONVERT_STAT_BITS_TXT_LF:
126 return "lf";
127 case CONVERT_STAT_BITS_TXT_CRLF:
128 return "crlf";
129 case CONVERT_STAT_BITS_TXT_LF | CONVERT_STAT_BITS_TXT_CRLF:
130 return "mixed";
131 default:
132 return "none";
133 }
134 }
135
136 const char *get_cached_convert_stats_ascii(const char *path)
137 {
138 const char *ret;
139 unsigned long sz;
140 void *data = read_blob_data_from_cache(path, &sz);
141 ret = gather_convert_stats_ascii(data, sz);
142 free(data);
143 return ret;
144 }
145
146 const char *get_wt_convert_stats_ascii(const char *path)
147 {
148 const char *ret = "";
149 struct strbuf sb = STRBUF_INIT;
150 if (strbuf_read_file(&sb, path, 0) >= 0)
151 ret = gather_convert_stats_ascii(sb.buf, sb.len);
152 strbuf_release(&sb);
153 return ret;
154 }
155
156 static int text_eol_is_crlf(void)
157 {
158 if (auto_crlf == AUTO_CRLF_TRUE)
159 return 1;
160 else if (auto_crlf == AUTO_CRLF_INPUT)
161 return 0;
162 if (core_eol == EOL_CRLF)
163 return 1;
164 if (core_eol == EOL_UNSET && EOL_NATIVE == EOL_CRLF)
165 return 1;
166 return 0;
167 }
168
169 static enum eol output_eol(enum crlf_action crlf_action)
170 {
171 switch (crlf_action) {
172 case CRLF_BINARY:
173 return EOL_UNSET;
174 case CRLF_TEXT_CRLF:
175 return EOL_CRLF;
176 case CRLF_TEXT_INPUT:
177 return EOL_LF;
178 case CRLF_UNDEFINED:
179 case CRLF_AUTO_CRLF:
180 return EOL_CRLF;
181 case CRLF_AUTO_INPUT:
182 return EOL_LF;
183 case CRLF_TEXT:
184 case CRLF_AUTO:
185 /* fall through */
186 return text_eol_is_crlf() ? EOL_CRLF : EOL_LF;
187 }
188 warning("Illegal crlf_action %d\n", (int)crlf_action);
189 return core_eol;
190 }
191
192 static void check_safe_crlf(const char *path, enum crlf_action crlf_action,
193 struct text_stat *old_stats, struct text_stat *new_stats,
194 enum safe_crlf checksafe)
195 {
196 if (old_stats->crlf && !new_stats->crlf ) {
197 /*
198 * CRLFs would not be restored by checkout
199 */
200 if (checksafe == SAFE_CRLF_WARN)
201 warning(_("CRLF will be replaced by LF in %s.\n"
202 "The file will have its original line"
203 " endings in your working directory."), path);
204 else /* i.e. SAFE_CRLF_FAIL */
205 die(_("CRLF would be replaced by LF in %s."), path);
206 } else if (old_stats->lonelf && !new_stats->lonelf ) {
207 /*
208 * CRLFs would be added by checkout
209 */
210 if (checksafe == SAFE_CRLF_WARN)
211 warning(_("LF will be replaced by CRLF in %s.\n"
212 "The file will have its original line"
213 " endings in your working directory."), path);
214 else /* i.e. SAFE_CRLF_FAIL */
215 die(_("LF would be replaced by CRLF in %s"), path);
216 }
217 }
218
219 static int has_cr_in_index(const char *path)
220 {
221 unsigned long sz;
222 void *data;
223 int has_cr;
224
225 data = read_blob_data_from_cache(path, &sz);
226 if (!data)
227 return 0;
228 has_cr = memchr(data, '\r', sz) != NULL;
229 free(data);
230 return has_cr;
231 }
232
233 static int will_convert_lf_to_crlf(size_t len, struct text_stat *stats,
234 enum crlf_action crlf_action)
235 {
236 if (output_eol(crlf_action) != EOL_CRLF)
237 return 0;
238 /* No "naked" LF? Nothing to convert, regardless. */
239 if (!stats->lonelf)
240 return 0;
241
242 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
243 /* If we have any CR or CRLF line endings, we do not touch it */
244 /* This is the new safer autocrlf-handling */
245 if (stats->lonecr || stats->crlf)
246 return 0;
247
248 if (convert_is_binary(len, stats))
249 return 0;
250 }
251 return 1;
252
253 }
254
255 static int crlf_to_git(const char *path, const char *src, size_t len,
256 struct strbuf *buf,
257 enum crlf_action crlf_action, enum safe_crlf checksafe)
258 {
259 struct text_stat stats;
260 char *dst;
261 int convert_crlf_into_lf;
262
263 if (crlf_action == CRLF_BINARY ||
264 (src && !len))
265 return 0;
266
267 /*
268 * If we are doing a dry-run and have no source buffer, there is
269 * nothing to analyze; we must assume we would convert.
270 */
271 if (!buf && !src)
272 return 1;
273
274 gather_stats(src, len, &stats);
275 /* Optimization: No CRLF? Nothing to convert, regardless. */
276 convert_crlf_into_lf = !!stats.crlf;
277
278 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
279 if (convert_is_binary(len, &stats))
280 return 0;
281 /*
282 * If the file in the index has any CR in it, do not
283 * convert. This is the new safer autocrlf handling,
284 * unless we want to renormalize in a merge or
285 * cherry-pick.
286 */
287 if ((checksafe != SAFE_CRLF_RENORMALIZE) && has_cr_in_index(path))
288 convert_crlf_into_lf = 0;
289 }
290 if ((checksafe == SAFE_CRLF_WARN ||
291 (checksafe == SAFE_CRLF_FAIL)) && len) {
292 struct text_stat new_stats;
293 memcpy(&new_stats, &stats, sizeof(new_stats));
294 /* simulate "git add" */
295 if (convert_crlf_into_lf) {
296 new_stats.lonelf += new_stats.crlf;
297 new_stats.crlf = 0;
298 }
299 /* simulate "git checkout" */
300 if (will_convert_lf_to_crlf(len, &new_stats, crlf_action)) {
301 new_stats.crlf += new_stats.lonelf;
302 new_stats.lonelf = 0;
303 }
304 check_safe_crlf(path, crlf_action, &stats, &new_stats, checksafe);
305 }
306 if (!convert_crlf_into_lf)
307 return 0;
308
309 /*
310 * At this point all of our source analysis is done, and we are sure we
311 * would convert. If we are in dry-run mode, we can give an answer.
312 */
313 if (!buf)
314 return 1;
315
316 /* only grow if not in place */
317 if (strbuf_avail(buf) + buf->len < len)
318 strbuf_grow(buf, len - buf->len);
319 dst = buf->buf;
320 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
321 /*
322 * If we guessed, we already know we rejected a file with
323 * lone CR, and we can strip a CR without looking at what
324 * follow it.
325 */
326 do {
327 unsigned char c = *src++;
328 if (c != '\r')
329 *dst++ = c;
330 } while (--len);
331 } else {
332 do {
333 unsigned char c = *src++;
334 if (! (c == '\r' && (1 < len && *src == '\n')))
335 *dst++ = c;
336 } while (--len);
337 }
338 strbuf_setlen(buf, dst - buf->buf);
339 return 1;
340 }
341
342 static int crlf_to_worktree(const char *path, const char *src, size_t len,
343 struct strbuf *buf, enum crlf_action crlf_action)
344 {
345 char *to_free = NULL;
346 struct text_stat stats;
347
348 if (!len || output_eol(crlf_action) != EOL_CRLF)
349 return 0;
350
351 gather_stats(src, len, &stats);
352 if (!will_convert_lf_to_crlf(len, &stats, crlf_action))
353 return 0;
354
355 /* are we "faking" in place editing ? */
356 if (src == buf->buf)
357 to_free = strbuf_detach(buf, NULL);
358
359 strbuf_grow(buf, len + stats.lonelf);
360 for (;;) {
361 const char *nl = memchr(src, '\n', len);
362 if (!nl)
363 break;
364 if (nl > src && nl[-1] == '\r') {
365 strbuf_add(buf, src, nl + 1 - src);
366 } else {
367 strbuf_add(buf, src, nl - src);
368 strbuf_addstr(buf, "\r\n");
369 }
370 len -= nl + 1 - src;
371 src = nl + 1;
372 }
373 strbuf_add(buf, src, len);
374
375 free(to_free);
376 return 1;
377 }
378
379 struct filter_params {
380 const char *src;
381 unsigned long size;
382 int fd;
383 const char *cmd;
384 const char *path;
385 };
386
387 static int filter_buffer_or_fd(int in, int out, void *data)
388 {
389 /*
390 * Spawn cmd and feed the buffer contents through its stdin.
391 */
392 struct child_process child_process = CHILD_PROCESS_INIT;
393 struct filter_params *params = (struct filter_params *)data;
394 int write_err, status;
395 const char *argv[] = { NULL, NULL };
396
397 /* apply % substitution to cmd */
398 struct strbuf cmd = STRBUF_INIT;
399 struct strbuf path = STRBUF_INIT;
400 struct strbuf_expand_dict_entry dict[] = {
401 { "f", NULL, },
402 { NULL, NULL, },
403 };
404
405 /* quote the path to preserve spaces, etc. */
406 sq_quote_buf(&path, params->path);
407 dict[0].value = path.buf;
408
409 /* expand all %f with the quoted path */
410 strbuf_expand(&cmd, params->cmd, strbuf_expand_dict_cb, &dict);
411 strbuf_release(&path);
412
413 argv[0] = cmd.buf;
414
415 child_process.argv = argv;
416 child_process.use_shell = 1;
417 child_process.in = -1;
418 child_process.out = out;
419
420 if (start_command(&child_process))
421 return error("cannot fork to run external filter '%s'", params->cmd);
422
423 sigchain_push(SIGPIPE, SIG_IGN);
424
425 if (params->src) {
426 write_err = (write_in_full(child_process.in,
427 params->src, params->size) < 0);
428 if (errno == EPIPE)
429 write_err = 0;
430 } else {
431 write_err = copy_fd(params->fd, child_process.in);
432 if (write_err == COPY_WRITE_ERROR && errno == EPIPE)
433 write_err = 0;
434 }
435
436 if (close(child_process.in))
437 write_err = 1;
438 if (write_err)
439 error("cannot feed the input to external filter '%s'", params->cmd);
440
441 sigchain_pop(SIGPIPE);
442
443 status = finish_command(&child_process);
444 if (status)
445 error("external filter '%s' failed %d", params->cmd, status);
446
447 strbuf_release(&cmd);
448 return (write_err || status);
449 }
450
451 static int apply_single_file_filter(const char *path, const char *src, size_t len, int fd,
452 struct strbuf *dst, const char *cmd)
453 {
454 /*
455 * Create a pipeline to have the command filter the buffer's
456 * contents.
457 *
458 * (child --> cmd) --> us
459 */
460 int err = 0;
461 struct strbuf nbuf = STRBUF_INIT;
462 struct async async;
463 struct filter_params params;
464
465 memset(&async, 0, sizeof(async));
466 async.proc = filter_buffer_or_fd;
467 async.data = &params;
468 async.out = -1;
469 params.src = src;
470 params.size = len;
471 params.fd = fd;
472 params.cmd = cmd;
473 params.path = path;
474
475 fflush(NULL);
476 if (start_async(&async))
477 return 0; /* error was already reported */
478
479 if (strbuf_read(&nbuf, async.out, len) < 0) {
480 err = error("read from external filter '%s' failed", cmd);
481 }
482 if (close(async.out)) {
483 err = error("read from external filter '%s' failed", cmd);
484 }
485 if (finish_async(&async)) {
486 err = error("external filter '%s' failed", cmd);
487 }
488
489 if (!err) {
490 strbuf_swap(dst, &nbuf);
491 }
492 strbuf_release(&nbuf);
493 return !err;
494 }
495
496 #define CAP_CLEAN (1u<<0)
497 #define CAP_SMUDGE (1u<<1)
498
499 struct subprocess_entry {
500 struct hashmap_entry ent; /* must be the first member! */
501 const char *cmd;
502 struct child_process process;
503 };
504
505 struct cmd2process {
506 struct subprocess_entry subprocess; /* must be the first member! */
507 unsigned int supported_capabilities;
508 };
509
510 static int cmd_process_map_initialized;
511 static struct hashmap cmd_process_map;
512
513 static int cmd2process_cmp(const struct subprocess_entry *e1,
514 const struct subprocess_entry *e2,
515 const void *unused)
516 {
517 return strcmp(e1->cmd, e2->cmd);
518 }
519
520 static struct subprocess_entry *find_multi_file_filter_entry(struct hashmap *hashmap, const char *cmd)
521 {
522 struct subprocess_entry key;
523
524 hashmap_entry_init(&key, strhash(cmd));
525 key.cmd = cmd;
526 return hashmap_get(hashmap, &key, NULL);
527 }
528
529 static void read_multi_file_filter_status(int fd, struct strbuf *status)
530 {
531 struct strbuf **pair;
532 char *line;
533 for (;;) {
534 line = packet_read_line(fd, NULL);
535 if (!line)
536 break;
537 pair = strbuf_split_str(line, '=', 2);
538 if (pair[0] && pair[0]->len && pair[1]) {
539 /* the last "status=<foo>" line wins */
540 if (!strcmp(pair[0]->buf, "status=")) {
541 strbuf_reset(status);
542 strbuf_addbuf(status, pair[1]);
543 }
544 }
545 strbuf_list_free(pair);
546 }
547 }
548
549 static void kill_multi_file_filter(struct hashmap *hashmap, struct subprocess_entry *entry)
550 {
551 if (!entry)
552 return;
553
554 entry->process.clean_on_exit = 0;
555 kill(entry->process.pid, SIGTERM);
556 finish_command(&entry->process);
557
558 hashmap_remove(hashmap, entry, NULL);
559 free(entry);
560 }
561
562 static void stop_multi_file_filter(struct child_process *process)
563 {
564 sigchain_push(SIGPIPE, SIG_IGN);
565 /* Closing the pipe signals the filter to initiate a shutdown. */
566 close(process->in);
567 close(process->out);
568 sigchain_pop(SIGPIPE);
569 /* Finish command will wait until the shutdown is complete. */
570 finish_command(process);
571 }
572
573 static int start_multi_file_filter_fn(struct cmd2process *entry)
574 {
575 int err;
576 struct string_list cap_list = STRING_LIST_INIT_NODUP;
577 char *cap_buf;
578 const char *cap_name;
579 struct child_process *process = &entry->subprocess.process;
580 const char *cmd = entry->subprocess.cmd;
581
582 sigchain_push(SIGPIPE, SIG_IGN);
583
584 err = packet_writel(process->in, "git-filter-client", "version=2", NULL);
585 if (err)
586 goto done;
587
588 err = strcmp(packet_read_line(process->out, NULL), "git-filter-server");
589 if (err) {
590 error("external filter '%s' does not support filter protocol version 2", cmd);
591 goto done;
592 }
593 err = strcmp(packet_read_line(process->out, NULL), "version=2");
594 if (err)
595 goto done;
596 err = packet_read_line(process->out, NULL) != NULL;
597 if (err)
598 goto done;
599
600 err = packet_writel(process->in, "capability=clean", "capability=smudge", NULL);
601
602 for (;;) {
603 cap_buf = packet_read_line(process->out, NULL);
604 if (!cap_buf)
605 break;
606 string_list_split_in_place(&cap_list, cap_buf, '=', 1);
607
608 if (cap_list.nr != 2 || strcmp(cap_list.items[0].string, "capability"))
609 continue;
610
611 cap_name = cap_list.items[1].string;
612 if (!strcmp(cap_name, "clean")) {
613 entry->supported_capabilities |= CAP_CLEAN;
614 } else if (!strcmp(cap_name, "smudge")) {
615 entry->supported_capabilities |= CAP_SMUDGE;
616 } else {
617 warning(
618 "external filter '%s' requested unsupported filter capability '%s'",
619 cmd, cap_name
620 );
621 }
622
623 string_list_clear(&cap_list, 0);
624 }
625
626 done:
627 sigchain_pop(SIGPIPE);
628
629 return err;
630 }
631
632 static struct cmd2process *start_multi_file_filter(struct hashmap *hashmap, const char *cmd)
633 {
634 int err;
635 struct cmd2process *entry;
636 struct child_process *process;
637 const char *argv[] = { cmd, NULL };
638
639 entry = xmalloc(sizeof(*entry));
640 entry->subprocess.cmd = cmd;
641 entry->supported_capabilities = 0;
642 process = &entry->subprocess.process;
643
644 child_process_init(process);
645 process->argv = argv;
646 process->use_shell = 1;
647 process->in = -1;
648 process->out = -1;
649 process->clean_on_exit = 1;
650 process->clean_on_exit_handler = stop_multi_file_filter;
651
652 if (start_command(process)) {
653 error("cannot fork to run external filter '%s'", cmd);
654 return NULL;
655 }
656
657 hashmap_entry_init(entry, strhash(cmd));
658
659 err = start_multi_file_filter_fn(entry);
660 if (err) {
661 error("initialization for external filter '%s' failed", cmd);
662 kill_multi_file_filter(hashmap, &entry->subprocess);
663 return NULL;
664 }
665
666 hashmap_add(hashmap, entry);
667 return entry;
668 }
669
670 static int apply_multi_file_filter(const char *path, const char *src, size_t len,
671 int fd, struct strbuf *dst, const char *cmd,
672 const unsigned int wanted_capability)
673 {
674 int err;
675 struct cmd2process *entry;
676 struct child_process *process;
677 struct strbuf nbuf = STRBUF_INIT;
678 struct strbuf filter_status = STRBUF_INIT;
679 const char *filter_type;
680
681 if (!cmd_process_map_initialized) {
682 cmd_process_map_initialized = 1;
683 hashmap_init(&cmd_process_map, (hashmap_cmp_fn) cmd2process_cmp, 0);
684 entry = NULL;
685 } else {
686 entry = (struct cmd2process *)find_multi_file_filter_entry(&cmd_process_map, cmd);
687 }
688
689 fflush(NULL);
690
691 if (!entry) {
692 entry = start_multi_file_filter(&cmd_process_map, cmd);
693 if (!entry)
694 return 0;
695 }
696 process = &entry->subprocess.process;
697
698 if (!(wanted_capability & entry->supported_capabilities))
699 return 0;
700
701 if (CAP_CLEAN & wanted_capability)
702 filter_type = "clean";
703 else if (CAP_SMUDGE & wanted_capability)
704 filter_type = "smudge";
705 else
706 die("unexpected filter type");
707
708 sigchain_push(SIGPIPE, SIG_IGN);
709
710 assert(strlen(filter_type) < LARGE_PACKET_DATA_MAX - strlen("command=\n"));
711 err = packet_write_fmt_gently(process->in, "command=%s\n", filter_type);
712 if (err)
713 goto done;
714
715 err = strlen(path) > LARGE_PACKET_DATA_MAX - strlen("pathname=\n");
716 if (err) {
717 error("path name too long for external filter");
718 goto done;
719 }
720
721 err = packet_write_fmt_gently(process->in, "pathname=%s\n", path);
722 if (err)
723 goto done;
724
725 err = packet_flush_gently(process->in);
726 if (err)
727 goto done;
728
729 if (fd >= 0)
730 err = write_packetized_from_fd(fd, process->in);
731 else
732 err = write_packetized_from_buf(src, len, process->in);
733 if (err)
734 goto done;
735
736 read_multi_file_filter_status(process->out, &filter_status);
737 err = strcmp(filter_status.buf, "success");
738 if (err)
739 goto done;
740
741 err = read_packetized_to_strbuf(process->out, &nbuf) < 0;
742 if (err)
743 goto done;
744
745 read_multi_file_filter_status(process->out, &filter_status);
746 err = strcmp(filter_status.buf, "success");
747
748 done:
749 sigchain_pop(SIGPIPE);
750
751 if (err) {
752 if (!strcmp(filter_status.buf, "error")) {
753 /* The filter signaled a problem with the file. */
754 } else if (!strcmp(filter_status.buf, "abort")) {
755 /*
756 * The filter signaled a permanent problem. Don't try to filter
757 * files with the same command for the lifetime of the current
758 * Git process.
759 */
760 entry->supported_capabilities &= ~wanted_capability;
761 } else {
762 /*
763 * Something went wrong with the protocol filter.
764 * Force shutdown and restart if another blob requires filtering.
765 */
766 error("external filter '%s' failed", cmd);
767 kill_multi_file_filter(&cmd_process_map, &entry->subprocess);
768 }
769 } else {
770 strbuf_swap(dst, &nbuf);
771 }
772 strbuf_release(&nbuf);
773 return !err;
774 }
775
776 static struct convert_driver {
777 const char *name;
778 struct convert_driver *next;
779 const char *smudge;
780 const char *clean;
781 const char *process;
782 int required;
783 } *user_convert, **user_convert_tail;
784
785 static int apply_filter(const char *path, const char *src, size_t len,
786 int fd, struct strbuf *dst, struct convert_driver *drv,
787 const unsigned int wanted_capability)
788 {
789 const char *cmd = NULL;
790
791 if (!drv)
792 return 0;
793
794 if (!dst)
795 return 1;
796
797 if ((CAP_CLEAN & wanted_capability) && !drv->process && drv->clean)
798 cmd = drv->clean;
799 else if ((CAP_SMUDGE & wanted_capability) && !drv->process && drv->smudge)
800 cmd = drv->smudge;
801
802 if (cmd && *cmd)
803 return apply_single_file_filter(path, src, len, fd, dst, cmd);
804 else if (drv->process && *drv->process)
805 return apply_multi_file_filter(path, src, len, fd, dst, drv->process, wanted_capability);
806
807 return 0;
808 }
809
810 static int read_convert_config(const char *var, const char *value, void *cb)
811 {
812 const char *key, *name;
813 int namelen;
814 struct convert_driver *drv;
815
816 /*
817 * External conversion drivers are configured using
818 * "filter.<name>.variable".
819 */
820 if (parse_config_key(var, "filter", &name, &namelen, &key) < 0 || !name)
821 return 0;
822 for (drv = user_convert; drv; drv = drv->next)
823 if (!strncmp(drv->name, name, namelen) && !drv->name[namelen])
824 break;
825 if (!drv) {
826 drv = xcalloc(1, sizeof(struct convert_driver));
827 drv->name = xmemdupz(name, namelen);
828 *user_convert_tail = drv;
829 user_convert_tail = &(drv->next);
830 }
831
832 /*
833 * filter.<name>.smudge and filter.<name>.clean specifies
834 * the command line:
835 *
836 * command-line
837 *
838 * The command-line will not be interpolated in any way.
839 */
840
841 if (!strcmp("smudge", key))
842 return git_config_string(&drv->smudge, var, value);
843
844 if (!strcmp("clean", key))
845 return git_config_string(&drv->clean, var, value);
846
847 if (!strcmp("process", key))
848 return git_config_string(&drv->process, var, value);
849
850 if (!strcmp("required", key)) {
851 drv->required = git_config_bool(var, value);
852 return 0;
853 }
854
855 return 0;
856 }
857
858 static int count_ident(const char *cp, unsigned long size)
859 {
860 /*
861 * "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$"
862 */
863 int cnt = 0;
864 char ch;
865
866 while (size) {
867 ch = *cp++;
868 size--;
869 if (ch != '$')
870 continue;
871 if (size < 3)
872 break;
873 if (memcmp("Id", cp, 2))
874 continue;
875 ch = cp[2];
876 cp += 3;
877 size -= 3;
878 if (ch == '$')
879 cnt++; /* $Id$ */
880 if (ch != ':')
881 continue;
882
883 /*
884 * "$Id: ... "; scan up to the closing dollar sign and discard.
885 */
886 while (size) {
887 ch = *cp++;
888 size--;
889 if (ch == '$') {
890 cnt++;
891 break;
892 }
893 if (ch == '\n')
894 break;
895 }
896 }
897 return cnt;
898 }
899
900 static int ident_to_git(const char *path, const char *src, size_t len,
901 struct strbuf *buf, int ident)
902 {
903 char *dst, *dollar;
904
905 if (!ident || (src && !count_ident(src, len)))
906 return 0;
907
908 if (!buf)
909 return 1;
910
911 /* only grow if not in place */
912 if (strbuf_avail(buf) + buf->len < len)
913 strbuf_grow(buf, len - buf->len);
914 dst = buf->buf;
915 for (;;) {
916 dollar = memchr(src, '$', len);
917 if (!dollar)
918 break;
919 memmove(dst, src, dollar + 1 - src);
920 dst += dollar + 1 - src;
921 len -= dollar + 1 - src;
922 src = dollar + 1;
923
924 if (len > 3 && !memcmp(src, "Id:", 3)) {
925 dollar = memchr(src + 3, '$', len - 3);
926 if (!dollar)
927 break;
928 if (memchr(src + 3, '\n', dollar - src - 3)) {
929 /* Line break before the next dollar. */
930 continue;
931 }
932
933 memcpy(dst, "Id$", 3);
934 dst += 3;
935 len -= dollar + 1 - src;
936 src = dollar + 1;
937 }
938 }
939 memmove(dst, src, len);
940 strbuf_setlen(buf, dst + len - buf->buf);
941 return 1;
942 }
943
944 static int ident_to_worktree(const char *path, const char *src, size_t len,
945 struct strbuf *buf, int ident)
946 {
947 unsigned char sha1[20];
948 char *to_free = NULL, *dollar, *spc;
949 int cnt;
950
951 if (!ident)
952 return 0;
953
954 cnt = count_ident(src, len);
955 if (!cnt)
956 return 0;
957
958 /* are we "faking" in place editing ? */
959 if (src == buf->buf)
960 to_free = strbuf_detach(buf, NULL);
961 hash_sha1_file(src, len, "blob", sha1);
962
963 strbuf_grow(buf, len + cnt * 43);
964 for (;;) {
965 /* step 1: run to the next '$' */
966 dollar = memchr(src, '$', len);
967 if (!dollar)
968 break;
969 strbuf_add(buf, src, dollar + 1 - src);
970 len -= dollar + 1 - src;
971 src = dollar + 1;
972
973 /* step 2: does it looks like a bit like Id:xxx$ or Id$ ? */
974 if (len < 3 || memcmp("Id", src, 2))
975 continue;
976
977 /* step 3: skip over Id$ or Id:xxxxx$ */
978 if (src[2] == '$') {
979 src += 3;
980 len -= 3;
981 } else if (src[2] == ':') {
982 /*
983 * It's possible that an expanded Id has crept its way into the
984 * repository, we cope with that by stripping the expansion out.
985 * This is probably not a good idea, since it will cause changes
986 * on checkout, which won't go away by stash, but let's keep it
987 * for git-style ids.
988 */
989 dollar = memchr(src + 3, '$', len - 3);
990 if (!dollar) {
991 /* incomplete keyword, no more '$', so just quit the loop */
992 break;
993 }
994
995 if (memchr(src + 3, '\n', dollar - src - 3)) {
996 /* Line break before the next dollar. */
997 continue;
998 }
999
1000 spc = memchr(src + 4, ' ', dollar - src - 4);
1001 if (spc && spc < dollar-1) {
1002 /* There are spaces in unexpected places.
1003 * This is probably an id from some other
1004 * versioning system. Keep it for now.
1005 */
1006 continue;
1007 }
1008
1009 len -= dollar + 1 - src;
1010 src = dollar + 1;
1011 } else {
1012 /* it wasn't a "Id$" or "Id:xxxx$" */
1013 continue;
1014 }
1015
1016 /* step 4: substitute */
1017 strbuf_addstr(buf, "Id: ");
1018 strbuf_add(buf, sha1_to_hex(sha1), 40);
1019 strbuf_addstr(buf, " $");
1020 }
1021 strbuf_add(buf, src, len);
1022
1023 free(to_free);
1024 return 1;
1025 }
1026
1027 static enum crlf_action git_path_check_crlf(struct attr_check_item *check)
1028 {
1029 const char *value = check->value;
1030
1031 if (ATTR_TRUE(value))
1032 return CRLF_TEXT;
1033 else if (ATTR_FALSE(value))
1034 return CRLF_BINARY;
1035 else if (ATTR_UNSET(value))
1036 ;
1037 else if (!strcmp(value, "input"))
1038 return CRLF_TEXT_INPUT;
1039 else if (!strcmp(value, "auto"))
1040 return CRLF_AUTO;
1041 return CRLF_UNDEFINED;
1042 }
1043
1044 static enum eol git_path_check_eol(struct attr_check_item *check)
1045 {
1046 const char *value = check->value;
1047
1048 if (ATTR_UNSET(value))
1049 ;
1050 else if (!strcmp(value, "lf"))
1051 return EOL_LF;
1052 else if (!strcmp(value, "crlf"))
1053 return EOL_CRLF;
1054 return EOL_UNSET;
1055 }
1056
1057 static struct convert_driver *git_path_check_convert(struct attr_check_item *check)
1058 {
1059 const char *value = check->value;
1060 struct convert_driver *drv;
1061
1062 if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
1063 return NULL;
1064 for (drv = user_convert; drv; drv = drv->next)
1065 if (!strcmp(value, drv->name))
1066 return drv;
1067 return NULL;
1068 }
1069
1070 static int git_path_check_ident(struct attr_check_item *check)
1071 {
1072 const char *value = check->value;
1073
1074 return !!ATTR_TRUE(value);
1075 }
1076
1077 struct conv_attrs {
1078 struct convert_driver *drv;
1079 enum crlf_action attr_action; /* What attr says */
1080 enum crlf_action crlf_action; /* When no attr is set, use core.autocrlf */
1081 int ident;
1082 };
1083
1084 static void convert_attrs(struct conv_attrs *ca, const char *path)
1085 {
1086 static struct attr_check *check;
1087
1088 if (!check) {
1089 check = attr_check_initl("crlf", "ident", "filter",
1090 "eol", "text", NULL);
1091 user_convert_tail = &user_convert;
1092 git_config(read_convert_config, NULL);
1093 }
1094
1095 if (!git_check_attr(path, check)) {
1096 struct attr_check_item *ccheck = check->items;
1097 ca->crlf_action = git_path_check_crlf(ccheck + 4);
1098 if (ca->crlf_action == CRLF_UNDEFINED)
1099 ca->crlf_action = git_path_check_crlf(ccheck + 0);
1100 ca->attr_action = ca->crlf_action;
1101 ca->ident = git_path_check_ident(ccheck + 1);
1102 ca->drv = git_path_check_convert(ccheck + 2);
1103 if (ca->crlf_action != CRLF_BINARY) {
1104 enum eol eol_attr = git_path_check_eol(ccheck + 3);
1105 if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_LF)
1106 ca->crlf_action = CRLF_AUTO_INPUT;
1107 else if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_CRLF)
1108 ca->crlf_action = CRLF_AUTO_CRLF;
1109 else if (eol_attr == EOL_LF)
1110 ca->crlf_action = CRLF_TEXT_INPUT;
1111 else if (eol_attr == EOL_CRLF)
1112 ca->crlf_action = CRLF_TEXT_CRLF;
1113 }
1114 ca->attr_action = ca->crlf_action;
1115 } else {
1116 ca->drv = NULL;
1117 ca->crlf_action = CRLF_UNDEFINED;
1118 ca->ident = 0;
1119 }
1120 if (ca->crlf_action == CRLF_TEXT)
1121 ca->crlf_action = text_eol_is_crlf() ? CRLF_TEXT_CRLF : CRLF_TEXT_INPUT;
1122 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_FALSE)
1123 ca->crlf_action = CRLF_BINARY;
1124 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_TRUE)
1125 ca->crlf_action = CRLF_AUTO_CRLF;
1126 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_INPUT)
1127 ca->crlf_action = CRLF_AUTO_INPUT;
1128 }
1129
1130 int would_convert_to_git_filter_fd(const char *path)
1131 {
1132 struct conv_attrs ca;
1133
1134 convert_attrs(&ca, path);
1135 if (!ca.drv)
1136 return 0;
1137
1138 /*
1139 * Apply a filter to an fd only if the filter is required to succeed.
1140 * We must die if the filter fails, because the original data before
1141 * filtering is not available.
1142 */
1143 if (!ca.drv->required)
1144 return 0;
1145
1146 return apply_filter(path, NULL, 0, -1, NULL, ca.drv, CAP_CLEAN);
1147 }
1148
1149 const char *get_convert_attr_ascii(const char *path)
1150 {
1151 struct conv_attrs ca;
1152
1153 convert_attrs(&ca, path);
1154 switch (ca.attr_action) {
1155 case CRLF_UNDEFINED:
1156 return "";
1157 case CRLF_BINARY:
1158 return "-text";
1159 case CRLF_TEXT:
1160 return "text";
1161 case CRLF_TEXT_INPUT:
1162 return "text eol=lf";
1163 case CRLF_TEXT_CRLF:
1164 return "text eol=crlf";
1165 case CRLF_AUTO:
1166 return "text=auto";
1167 case CRLF_AUTO_CRLF:
1168 return "text=auto eol=crlf";
1169 case CRLF_AUTO_INPUT:
1170 return "text=auto eol=lf";
1171 }
1172 return "";
1173 }
1174
1175 int convert_to_git(const char *path, const char *src, size_t len,
1176 struct strbuf *dst, enum safe_crlf checksafe)
1177 {
1178 int ret = 0;
1179 struct conv_attrs ca;
1180
1181 convert_attrs(&ca, path);
1182
1183 ret |= apply_filter(path, src, len, -1, dst, ca.drv, CAP_CLEAN);
1184 if (!ret && ca.drv && ca.drv->required)
1185 die("%s: clean filter '%s' failed", path, ca.drv->name);
1186
1187 if (ret && dst) {
1188 src = dst->buf;
1189 len = dst->len;
1190 }
1191 ret |= crlf_to_git(path, src, len, dst, ca.crlf_action, checksafe);
1192 if (ret && dst) {
1193 src = dst->buf;
1194 len = dst->len;
1195 }
1196 return ret | ident_to_git(path, src, len, dst, ca.ident);
1197 }
1198
1199 void convert_to_git_filter_fd(const char *path, int fd, struct strbuf *dst,
1200 enum safe_crlf checksafe)
1201 {
1202 struct conv_attrs ca;
1203 convert_attrs(&ca, path);
1204
1205 assert(ca.drv);
1206 assert(ca.drv->clean || ca.drv->process);
1207
1208 if (!apply_filter(path, NULL, 0, fd, dst, ca.drv, CAP_CLEAN))
1209 die("%s: clean filter '%s' failed", path, ca.drv->name);
1210
1211 crlf_to_git(path, dst->buf, dst->len, dst, ca.crlf_action, checksafe);
1212 ident_to_git(path, dst->buf, dst->len, dst, ca.ident);
1213 }
1214
1215 static int convert_to_working_tree_internal(const char *path, const char *src,
1216 size_t len, struct strbuf *dst,
1217 int normalizing)
1218 {
1219 int ret = 0, ret_filter = 0;
1220 struct conv_attrs ca;
1221
1222 convert_attrs(&ca, path);
1223
1224 ret |= ident_to_worktree(path, src, len, dst, ca.ident);
1225 if (ret) {
1226 src = dst->buf;
1227 len = dst->len;
1228 }
1229 /*
1230 * CRLF conversion can be skipped if normalizing, unless there
1231 * is a smudge or process filter (even if the process filter doesn't
1232 * support smudge). The filters might expect CRLFs.
1233 */
1234 if ((ca.drv && (ca.drv->smudge || ca.drv->process)) || !normalizing) {
1235 ret |= crlf_to_worktree(path, src, len, dst, ca.crlf_action);
1236 if (ret) {
1237 src = dst->buf;
1238 len = dst->len;
1239 }
1240 }
1241
1242 ret_filter = apply_filter(path, src, len, -1, dst, ca.drv, CAP_SMUDGE);
1243 if (!ret_filter && ca.drv && ca.drv->required)
1244 die("%s: smudge filter %s failed", path, ca.drv->name);
1245
1246 return ret | ret_filter;
1247 }
1248
1249 int convert_to_working_tree(const char *path, const char *src, size_t len, struct strbuf *dst)
1250 {
1251 return convert_to_working_tree_internal(path, src, len, dst, 0);
1252 }
1253
1254 int renormalize_buffer(const char *path, const char *src, size_t len, struct strbuf *dst)
1255 {
1256 int ret = convert_to_working_tree_internal(path, src, len, dst, 1);
1257 if (ret) {
1258 src = dst->buf;
1259 len = dst->len;
1260 }
1261 return ret | convert_to_git(path, src, len, dst, SAFE_CRLF_RENORMALIZE);
1262 }
1263
1264 /*****************************************************************
1265 *
1266 * Streaming conversion support
1267 *
1268 *****************************************************************/
1269
1270 typedef int (*filter_fn)(struct stream_filter *,
1271 const char *input, size_t *isize_p,
1272 char *output, size_t *osize_p);
1273 typedef void (*free_fn)(struct stream_filter *);
1274
1275 struct stream_filter_vtbl {
1276 filter_fn filter;
1277 free_fn free;
1278 };
1279
1280 struct stream_filter {
1281 struct stream_filter_vtbl *vtbl;
1282 };
1283
1284 static int null_filter_fn(struct stream_filter *filter,
1285 const char *input, size_t *isize_p,
1286 char *output, size_t *osize_p)
1287 {
1288 size_t count;
1289
1290 if (!input)
1291 return 0; /* we do not keep any states */
1292 count = *isize_p;
1293 if (*osize_p < count)
1294 count = *osize_p;
1295 if (count) {
1296 memmove(output, input, count);
1297 *isize_p -= count;
1298 *osize_p -= count;
1299 }
1300 return 0;
1301 }
1302
1303 static void null_free_fn(struct stream_filter *filter)
1304 {
1305 ; /* nothing -- null instances are shared */
1306 }
1307
1308 static struct stream_filter_vtbl null_vtbl = {
1309 null_filter_fn,
1310 null_free_fn,
1311 };
1312
1313 static struct stream_filter null_filter_singleton = {
1314 &null_vtbl,
1315 };
1316
1317 int is_null_stream_filter(struct stream_filter *filter)
1318 {
1319 return filter == &null_filter_singleton;
1320 }
1321
1322
1323 /*
1324 * LF-to-CRLF filter
1325 */
1326
1327 struct lf_to_crlf_filter {
1328 struct stream_filter filter;
1329 unsigned has_held:1;
1330 char held;
1331 };
1332
1333 static int lf_to_crlf_filter_fn(struct stream_filter *filter,
1334 const char *input, size_t *isize_p,
1335 char *output, size_t *osize_p)
1336 {
1337 size_t count, o = 0;
1338 struct lf_to_crlf_filter *lf_to_crlf = (struct lf_to_crlf_filter *)filter;
1339
1340 /*
1341 * We may be holding onto the CR to see if it is followed by a
1342 * LF, in which case we would need to go to the main loop.
1343 * Otherwise, just emit it to the output stream.
1344 */
1345 if (lf_to_crlf->has_held && (lf_to_crlf->held != '\r' || !input)) {
1346 output[o++] = lf_to_crlf->held;
1347 lf_to_crlf->has_held = 0;
1348 }
1349
1350 /* We are told to drain */
1351 if (!input) {
1352 *osize_p -= o;
1353 return 0;
1354 }
1355
1356 count = *isize_p;
1357 if (count || lf_to_crlf->has_held) {
1358 size_t i;
1359 int was_cr = 0;
1360
1361 if (lf_to_crlf->has_held) {
1362 was_cr = 1;
1363 lf_to_crlf->has_held = 0;
1364 }
1365
1366 for (i = 0; o < *osize_p && i < count; i++) {
1367 char ch = input[i];
1368
1369 if (ch == '\n') {
1370 output[o++] = '\r';
1371 } else if (was_cr) {
1372 /*
1373 * Previous round saw CR and it is not followed
1374 * by a LF; emit the CR before processing the
1375 * current character.
1376 */
1377 output[o++] = '\r';
1378 }
1379
1380 /*
1381 * We may have consumed the last output slot,
1382 * in which case we need to break out of this
1383 * loop; hold the current character before
1384 * returning.
1385 */
1386 if (*osize_p <= o) {
1387 lf_to_crlf->has_held = 1;
1388 lf_to_crlf->held = ch;
1389 continue; /* break but increment i */
1390 }
1391
1392 if (ch == '\r') {
1393 was_cr = 1;
1394 continue;
1395 }
1396
1397 was_cr = 0;
1398 output[o++] = ch;
1399 }
1400
1401 *osize_p -= o;
1402 *isize_p -= i;
1403
1404 if (!lf_to_crlf->has_held && was_cr) {
1405 lf_to_crlf->has_held = 1;
1406 lf_to_crlf->held = '\r';
1407 }
1408 }
1409 return 0;
1410 }
1411
1412 static void lf_to_crlf_free_fn(struct stream_filter *filter)
1413 {
1414 free(filter);
1415 }
1416
1417 static struct stream_filter_vtbl lf_to_crlf_vtbl = {
1418 lf_to_crlf_filter_fn,
1419 lf_to_crlf_free_fn,
1420 };
1421
1422 static struct stream_filter *lf_to_crlf_filter(void)
1423 {
1424 struct lf_to_crlf_filter *lf_to_crlf = xcalloc(1, sizeof(*lf_to_crlf));
1425
1426 lf_to_crlf->filter.vtbl = &lf_to_crlf_vtbl;
1427 return (struct stream_filter *)lf_to_crlf;
1428 }
1429
1430 /*
1431 * Cascade filter
1432 */
1433 #define FILTER_BUFFER 1024
1434 struct cascade_filter {
1435 struct stream_filter filter;
1436 struct stream_filter *one;
1437 struct stream_filter *two;
1438 char buf[FILTER_BUFFER];
1439 int end, ptr;
1440 };
1441
1442 static int cascade_filter_fn(struct stream_filter *filter,
1443 const char *input, size_t *isize_p,
1444 char *output, size_t *osize_p)
1445 {
1446 struct cascade_filter *cas = (struct cascade_filter *) filter;
1447 size_t filled = 0;
1448 size_t sz = *osize_p;
1449 size_t to_feed, remaining;
1450
1451 /*
1452 * input -- (one) --> buf -- (two) --> output
1453 */
1454 while (filled < sz) {
1455 remaining = sz - filled;
1456
1457 /* do we already have something to feed two with? */
1458 if (cas->ptr < cas->end) {
1459 to_feed = cas->end - cas->ptr;
1460 if (stream_filter(cas->two,
1461 cas->buf + cas->ptr, &to_feed,
1462 output + filled, &remaining))
1463 return -1;
1464 cas->ptr += (cas->end - cas->ptr) - to_feed;
1465 filled = sz - remaining;
1466 continue;
1467 }
1468
1469 /* feed one from upstream and have it emit into our buffer */
1470 to_feed = input ? *isize_p : 0;
1471 if (input && !to_feed)
1472 break;
1473 remaining = sizeof(cas->buf);
1474 if (stream_filter(cas->one,
1475 input, &to_feed,
1476 cas->buf, &remaining))
1477 return -1;
1478 cas->end = sizeof(cas->buf) - remaining;
1479 cas->ptr = 0;
1480 if (input) {
1481 size_t fed = *isize_p - to_feed;
1482 *isize_p -= fed;
1483 input += fed;
1484 }
1485
1486 /* do we know that we drained one completely? */
1487 if (input || cas->end)
1488 continue;
1489
1490 /* tell two to drain; we have nothing more to give it */
1491 to_feed = 0;
1492 remaining = sz - filled;
1493 if (stream_filter(cas->two,
1494 NULL, &to_feed,
1495 output + filled, &remaining))
1496 return -1;
1497 if (remaining == (sz - filled))
1498 break; /* completely drained two */
1499 filled = sz - remaining;
1500 }
1501 *osize_p -= filled;
1502 return 0;
1503 }
1504
1505 static void cascade_free_fn(struct stream_filter *filter)
1506 {
1507 struct cascade_filter *cas = (struct cascade_filter *)filter;
1508 free_stream_filter(cas->one);
1509 free_stream_filter(cas->two);
1510 free(filter);
1511 }
1512
1513 static struct stream_filter_vtbl cascade_vtbl = {
1514 cascade_filter_fn,
1515 cascade_free_fn,
1516 };
1517
1518 static struct stream_filter *cascade_filter(struct stream_filter *one,
1519 struct stream_filter *two)
1520 {
1521 struct cascade_filter *cascade;
1522
1523 if (!one || is_null_stream_filter(one))
1524 return two;
1525 if (!two || is_null_stream_filter(two))
1526 return one;
1527
1528 cascade = xmalloc(sizeof(*cascade));
1529 cascade->one = one;
1530 cascade->two = two;
1531 cascade->end = cascade->ptr = 0;
1532 cascade->filter.vtbl = &cascade_vtbl;
1533 return (struct stream_filter *)cascade;
1534 }
1535
1536 /*
1537 * ident filter
1538 */
1539 #define IDENT_DRAINING (-1)
1540 #define IDENT_SKIPPING (-2)
1541 struct ident_filter {
1542 struct stream_filter filter;
1543 struct strbuf left;
1544 int state;
1545 char ident[45]; /* ": x40 $" */
1546 };
1547
1548 static int is_foreign_ident(const char *str)
1549 {
1550 int i;
1551
1552 if (!skip_prefix(str, "$Id: ", &str))
1553 return 0;
1554 for (i = 0; str[i]; i++) {
1555 if (isspace(str[i]) && str[i+1] != '$')
1556 return 1;
1557 }
1558 return 0;
1559 }
1560
1561 static void ident_drain(struct ident_filter *ident, char **output_p, size_t *osize_p)
1562 {
1563 size_t to_drain = ident->left.len;
1564
1565 if (*osize_p < to_drain)
1566 to_drain = *osize_p;
1567 if (to_drain) {
1568 memcpy(*output_p, ident->left.buf, to_drain);
1569 strbuf_remove(&ident->left, 0, to_drain);
1570 *output_p += to_drain;
1571 *osize_p -= to_drain;
1572 }
1573 if (!ident->left.len)
1574 ident->state = 0;
1575 }
1576
1577 static int ident_filter_fn(struct stream_filter *filter,
1578 const char *input, size_t *isize_p,
1579 char *output, size_t *osize_p)
1580 {
1581 struct ident_filter *ident = (struct ident_filter *)filter;
1582 static const char head[] = "$Id";
1583
1584 if (!input) {
1585 /* drain upon eof */
1586 switch (ident->state) {
1587 default:
1588 strbuf_add(&ident->left, head, ident->state);
1589 case IDENT_SKIPPING:
1590 /* fallthru */
1591 case IDENT_DRAINING:
1592 ident_drain(ident, &output, osize_p);
1593 }
1594 return 0;
1595 }
1596
1597 while (*isize_p || (ident->state == IDENT_DRAINING)) {
1598 int ch;
1599
1600 if (ident->state == IDENT_DRAINING) {
1601 ident_drain(ident, &output, osize_p);
1602 if (!*osize_p)
1603 break;
1604 continue;
1605 }
1606
1607 ch = *(input++);
1608 (*isize_p)--;
1609
1610 if (ident->state == IDENT_SKIPPING) {
1611 /*
1612 * Skipping until '$' or LF, but keeping them
1613 * in case it is a foreign ident.
1614 */
1615 strbuf_addch(&ident->left, ch);
1616 if (ch != '\n' && ch != '$')
1617 continue;
1618 if (ch == '$' && !is_foreign_ident(ident->left.buf)) {
1619 strbuf_setlen(&ident->left, sizeof(head) - 1);
1620 strbuf_addstr(&ident->left, ident->ident);
1621 }
1622 ident->state = IDENT_DRAINING;
1623 continue;
1624 }
1625
1626 if (ident->state < sizeof(head) &&
1627 head[ident->state] == ch) {
1628 ident->state++;
1629 continue;
1630 }
1631
1632 if (ident->state)
1633 strbuf_add(&ident->left, head, ident->state);
1634 if (ident->state == sizeof(head) - 1) {
1635 if (ch != ':' && ch != '$') {
1636 strbuf_addch(&ident->left, ch);
1637 ident->state = 0;
1638 continue;
1639 }
1640
1641 if (ch == ':') {
1642 strbuf_addch(&ident->left, ch);
1643 ident->state = IDENT_SKIPPING;
1644 } else {
1645 strbuf_addstr(&ident->left, ident->ident);
1646 ident->state = IDENT_DRAINING;
1647 }
1648 continue;
1649 }
1650
1651 strbuf_addch(&ident->left, ch);
1652 ident->state = IDENT_DRAINING;
1653 }
1654 return 0;
1655 }
1656
1657 static void ident_free_fn(struct stream_filter *filter)
1658 {
1659 struct ident_filter *ident = (struct ident_filter *)filter;
1660 strbuf_release(&ident->left);
1661 free(filter);
1662 }
1663
1664 static struct stream_filter_vtbl ident_vtbl = {
1665 ident_filter_fn,
1666 ident_free_fn,
1667 };
1668
1669 static struct stream_filter *ident_filter(const unsigned char *sha1)
1670 {
1671 struct ident_filter *ident = xmalloc(sizeof(*ident));
1672
1673 xsnprintf(ident->ident, sizeof(ident->ident),
1674 ": %s $", sha1_to_hex(sha1));
1675 strbuf_init(&ident->left, 0);
1676 ident->filter.vtbl = &ident_vtbl;
1677 ident->state = 0;
1678 return (struct stream_filter *)ident;
1679 }
1680
1681 /*
1682 * Return an appropriately constructed filter for the path, or NULL if
1683 * the contents cannot be filtered without reading the whole thing
1684 * in-core.
1685 *
1686 * Note that you would be crazy to set CRLF, smuge/clean or ident to a
1687 * large binary blob you would want us not to slurp into the memory!
1688 */
1689 struct stream_filter *get_stream_filter(const char *path, const unsigned char *sha1)
1690 {
1691 struct conv_attrs ca;
1692 struct stream_filter *filter = NULL;
1693
1694 convert_attrs(&ca, path);
1695 if (ca.drv && (ca.drv->process || ca.drv->smudge || ca.drv->clean))
1696 return NULL;
1697
1698 if (ca.crlf_action == CRLF_AUTO || ca.crlf_action == CRLF_AUTO_CRLF)
1699 return NULL;
1700
1701 if (ca.ident)
1702 filter = ident_filter(sha1);
1703
1704 if (output_eol(ca.crlf_action) == EOL_CRLF)
1705 filter = cascade_filter(filter, lf_to_crlf_filter());
1706 else
1707 filter = cascade_filter(filter, &null_filter_singleton);
1708
1709 return filter;
1710 }
1711
1712 void free_stream_filter(struct stream_filter *filter)
1713 {
1714 filter->vtbl->free(filter);
1715 }
1716
1717 int stream_filter(struct stream_filter *filter,
1718 const char *input, size_t *isize_p,
1719 char *output, size_t *osize_p)
1720 {
1721 return filter->vtbl->filter(filter, input, isize_p, output, osize_p);
1722 }