?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
fsck: tighten error-checks of "git fsck <head>"
[git/git-svn.git] / convert.c
blobbe91358462a72c9d5a6c02f2b0e44585a76d4783
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 convert.
283 * This is the new safer autocrlf handling.
284 */
285 if (checksafe == SAFE_CRLF_RENORMALIZE)
286 checksafe = SAFE_CRLF_FALSE;
287 else if (has_cr_in_index(path))
288 convert_crlf_into_lf = 0;
289 }
290 if (checksafe && len) {
291 struct text_stat new_stats;
292 memcpy(&new_stats, &stats, sizeof(new_stats));
293 /* simulate "git add" */
294 if (convert_crlf_into_lf) {
295 new_stats.lonelf += new_stats.crlf;
296 new_stats.crlf = 0;
297 }
298 /* simulate "git checkout" */
299 if (will_convert_lf_to_crlf(len, &new_stats, crlf_action)) {
300 new_stats.crlf += new_stats.lonelf;
301 new_stats.lonelf = 0;
302 }
303 check_safe_crlf(path, crlf_action, &stats, &new_stats, checksafe);
304 }
305 if (!convert_crlf_into_lf)
306 return 0;
307
308 /*
309 * At this point all of our source analysis is done, and we are sure we
310 * would convert. If we are in dry-run mode, we can give an answer.
311 */
312 if (!buf)
313 return 1;
314
315 /* only grow if not in place */
316 if (strbuf_avail(buf) + buf->len < len)
317 strbuf_grow(buf, len - buf->len);
318 dst = buf->buf;
319 if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) {
320 /*
321 * If we guessed, we already know we rejected a file with
322 * lone CR, and we can strip a CR without looking at what
323 * follow it.
324 */
325 do {
326 unsigned char c = *src++;
327 if (c != '\r')
328 *dst++ = c;
329 } while (--len);
330 } else {
331 do {
332 unsigned char c = *src++;
333 if (! (c == '\r' && (1 < len && *src == '\n')))
334 *dst++ = c;
335 } while (--len);
336 }
337 strbuf_setlen(buf, dst - buf->buf);
338 return 1;
339 }
340
341 static int crlf_to_worktree(const char *path, const char *src, size_t len,
342 struct strbuf *buf, enum crlf_action crlf_action)
343 {
344 char *to_free = NULL;
345 struct text_stat stats;
346
347 if (!len || output_eol(crlf_action) != EOL_CRLF)
348 return 0;
349
350 gather_stats(src, len, &stats);
351 if (!will_convert_lf_to_crlf(len, &stats, crlf_action))
352 return 0;
353
354 /* are we "faking" in place editing ? */
355 if (src == buf->buf)
356 to_free = strbuf_detach(buf, NULL);
357
358 strbuf_grow(buf, len + stats.lonelf);
359 for (;;) {
360 const char *nl = memchr(src, '\n', len);
361 if (!nl)
362 break;
363 if (nl > src && nl[-1] == '\r') {
364 strbuf_add(buf, src, nl + 1 - src);
365 } else {
366 strbuf_add(buf, src, nl - src);
367 strbuf_addstr(buf, "\r\n");
368 }
369 len -= nl + 1 - src;
370 src = nl + 1;
371 }
372 strbuf_add(buf, src, len);
373
374 free(to_free);
375 return 1;
376 }
377
378 struct filter_params {
379 const char *src;
380 unsigned long size;
381 int fd;
382 const char *cmd;
383 const char *path;
384 };
385
386 static int filter_buffer_or_fd(int in, int out, void *data)
387 {
388 /*
389 * Spawn cmd and feed the buffer contents through its stdin.
390 */
391 struct child_process child_process = CHILD_PROCESS_INIT;
392 struct filter_params *params = (struct filter_params *)data;
393 int write_err, status;
394 const char *argv[] = { NULL, NULL };
395
396 /* apply % substitution to cmd */
397 struct strbuf cmd = STRBUF_INIT;
398 struct strbuf path = STRBUF_INIT;
399 struct strbuf_expand_dict_entry dict[] = {
400 { "f", NULL, },
401 { NULL, NULL, },
402 };
403
404 /* quote the path to preserve spaces, etc. */
405 sq_quote_buf(&path, params->path);
406 dict[0].value = path.buf;
407
408 /* expand all %f with the quoted path */
409 strbuf_expand(&cmd, params->cmd, strbuf_expand_dict_cb, &dict);
410 strbuf_release(&path);
411
412 argv[0] = cmd.buf;
413
414 child_process.argv = argv;
415 child_process.use_shell = 1;
416 child_process.in = -1;
417 child_process.out = out;
418
419 if (start_command(&child_process))
420 return error("cannot fork to run external filter '%s'", params->cmd);
421
422 sigchain_push(SIGPIPE, SIG_IGN);
423
424 if (params->src) {
425 write_err = (write_in_full(child_process.in,
426 params->src, params->size) < 0);
427 if (errno == EPIPE)
428 write_err = 0;
429 } else {
430 write_err = copy_fd(params->fd, child_process.in);
431 if (write_err == COPY_WRITE_ERROR && errno == EPIPE)
432 write_err = 0;
433 }
434
435 if (close(child_process.in))
436 write_err = 1;
437 if (write_err)
438 error("cannot feed the input to external filter '%s'", params->cmd);
439
440 sigchain_pop(SIGPIPE);
441
442 status = finish_command(&child_process);
443 if (status)
444 error("external filter '%s' failed %d", params->cmd, status);
445
446 strbuf_release(&cmd);
447 return (write_err || status);
448 }
449
450 static int apply_single_file_filter(const char *path, const char *src, size_t len, int fd,
451 struct strbuf *dst, const char *cmd)
452 {
453 /*
454 * Create a pipeline to have the command filter the buffer's
455 * contents.
456 *
457 * (child --> cmd) --> us
458 */
459 int err = 0;
460 struct strbuf nbuf = STRBUF_INIT;
461 struct async async;
462 struct filter_params params;
463
464 memset(&async, 0, sizeof(async));
465 async.proc = filter_buffer_or_fd;
466 async.data = &params;
467 async.out = -1;
468 params.src = src;
469 params.size = len;
470 params.fd = fd;
471 params.cmd = cmd;
472 params.path = path;
473
474 fflush(NULL);
475 if (start_async(&async))
476 return 0; /* error was already reported */
477
478 if (strbuf_read(&nbuf, async.out, len) < 0) {
479 err = error("read from external filter '%s' failed", cmd);
480 }
481 if (close(async.out)) {
482 err = error("read from external filter '%s' failed", cmd);
483 }
484 if (finish_async(&async)) {
485 err = error("external filter '%s' failed", cmd);
486 }
487
488 if (!err) {
489 strbuf_swap(dst, &nbuf);
490 }
491 strbuf_release(&nbuf);
492 return !err;
493 }
494
495 #define CAP_CLEAN (1u<<0)
496 #define CAP_SMUDGE (1u<<1)
497
498 struct cmd2process {
499 struct hashmap_entry ent; /* must be the first member! */
500 unsigned int supported_capabilities;
501 const char *cmd;
502 struct child_process process;
503 };
504
505 static int cmd_process_map_initialized;
506 static struct hashmap cmd_process_map;
507
508 static int cmd2process_cmp(const struct cmd2process *e1,
509 const struct cmd2process *e2,
510 const void *unused)
511 {
512 return strcmp(e1->cmd, e2->cmd);
513 }
514
515 static struct cmd2process *find_multi_file_filter_entry(struct hashmap *hashmap, const char *cmd)
516 {
517 struct cmd2process key;
518 hashmap_entry_init(&key, strhash(cmd));
519 key.cmd = cmd;
520 return hashmap_get(hashmap, &key, NULL);
521 }
522
523 static int packet_write_list(int fd, const char *line, ...)
524 {
525 va_list args;
526 int err;
527 va_start(args, line);
528 for (;;) {
529 if (!line)
530 break;
531 if (strlen(line) > LARGE_PACKET_DATA_MAX)
532 return -1;
533 err = packet_write_fmt_gently(fd, "%s\n", line);
534 if (err)
535 return err;
536 line = va_arg(args, const char*);
537 }
538 va_end(args);
539 return packet_flush_gently(fd);
540 }
541
542 static void read_multi_file_filter_status(int fd, struct strbuf *status)
543 {
544 struct strbuf **pair;
545 char *line;
546 for (;;) {
547 line = packet_read_line(fd, NULL);
548 if (!line)
549 break;
550 pair = strbuf_split_str(line, '=', 2);
551 if (pair[0] && pair[0]->len && pair[1]) {
552 /* the last "status=<foo>" line wins */
553 if (!strcmp(pair[0]->buf, "status=")) {
554 strbuf_reset(status);
555 strbuf_addbuf(status, pair[1]);
556 }
557 }
558 strbuf_list_free(pair);
559 }
560 }
561
562 static void kill_multi_file_filter(struct hashmap *hashmap, struct cmd2process *entry)
563 {
564 if (!entry)
565 return;
566
567 entry->process.clean_on_exit = 0;
568 kill(entry->process.pid, SIGTERM);
569 finish_command(&entry->process);
570
571 hashmap_remove(hashmap, entry, NULL);
572 free(entry);
573 }
574
575 static void stop_multi_file_filter(struct child_process *process)
576 {
577 sigchain_push(SIGPIPE, SIG_IGN);
578 /* Closing the pipe signals the filter to initiate a shutdown. */
579 close(process->in);
580 close(process->out);
581 sigchain_pop(SIGPIPE);
582 /* Finish command will wait until the shutdown is complete. */
583 finish_command(process);
584 }
585
586 static struct cmd2process *start_multi_file_filter(struct hashmap *hashmap, const char *cmd)
587 {
588 int err;
589 struct cmd2process *entry;
590 struct child_process *process;
591 const char *argv[] = { cmd, NULL };
592 struct string_list cap_list = STRING_LIST_INIT_NODUP;
593 char *cap_buf;
594 const char *cap_name;
595
596 entry = xmalloc(sizeof(*entry));
597 entry->cmd = cmd;
598 entry->supported_capabilities = 0;
599 process = &entry->process;
600
601 child_process_init(process);
602 process->argv = argv;
603 process->use_shell = 1;
604 process->in = -1;
605 process->out = -1;
606 process->clean_on_exit = 1;
607 process->clean_on_exit_handler = stop_multi_file_filter;
608
609 if (start_command(process)) {
610 error("cannot fork to run external filter '%s'", cmd);
611 return NULL;
612 }
613
614 hashmap_entry_init(entry, strhash(cmd));
615
616 sigchain_push(SIGPIPE, SIG_IGN);
617
618 err = packet_write_list(process->in, "git-filter-client", "version=2", NULL);
619 if (err)
620 goto done;
621
622 err = strcmp(packet_read_line(process->out, NULL), "git-filter-server");
623 if (err) {
624 error("external filter '%s' does not support filter protocol version 2", cmd);
625 goto done;
626 }
627 err = strcmp(packet_read_line(process->out, NULL), "version=2");
628 if (err)
629 goto done;
630 err = packet_read_line(process->out, NULL) != NULL;
631 if (err)
632 goto done;
633
634 err = packet_write_list(process->in, "capability=clean", "capability=smudge", NULL);
635
636 for (;;) {
637 cap_buf = packet_read_line(process->out, NULL);
638 if (!cap_buf)
639 break;
640 string_list_split_in_place(&cap_list, cap_buf, '=', 1);
641
642 if (cap_list.nr != 2 || strcmp(cap_list.items[0].string, "capability"))
643 continue;
644
645 cap_name = cap_list.items[1].string;
646 if (!strcmp(cap_name, "clean")) {
647 entry->supported_capabilities |= CAP_CLEAN;
648 } else if (!strcmp(cap_name, "smudge")) {
649 entry->supported_capabilities |= CAP_SMUDGE;
650 } else {
651 warning(
652 "external filter '%s' requested unsupported filter capability '%s'",
653 cmd, cap_name
654 );
655 }
656
657 string_list_clear(&cap_list, 0);
658 }
659
660 done:
661 sigchain_pop(SIGPIPE);
662
663 if (err || errno == EPIPE) {
664 error("initialization for external filter '%s' failed", cmd);
665 kill_multi_file_filter(hashmap, entry);
666 return NULL;
667 }
668
669 hashmap_add(hashmap, entry);
670 return entry;
671 }
672
673 static int apply_multi_file_filter(const char *path, const char *src, size_t len,
674 int fd, struct strbuf *dst, const char *cmd,
675 const unsigned int wanted_capability)
676 {
677 int err;
678 struct cmd2process *entry;
679 struct child_process *process;
680 struct strbuf nbuf = STRBUF_INIT;
681 struct strbuf filter_status = STRBUF_INIT;
682 const char *filter_type;
683
684 if (!cmd_process_map_initialized) {
685 cmd_process_map_initialized = 1;
686 hashmap_init(&cmd_process_map, (hashmap_cmp_fn) cmd2process_cmp, 0);
687 entry = NULL;
688 } else {
689 entry = find_multi_file_filter_entry(&cmd_process_map, cmd);
690 }
691
692 fflush(NULL);
693
694 if (!entry) {
695 entry = start_multi_file_filter(&cmd_process_map, cmd);
696 if (!entry)
697 return 0;
698 }
699 process = &entry->process;
700
701 if (!(wanted_capability & entry->supported_capabilities))
702 return 0;
703
704 if (CAP_CLEAN & wanted_capability)
705 filter_type = "clean";
706 else if (CAP_SMUDGE & wanted_capability)
707 filter_type = "smudge";
708 else
709 die("unexpected filter type");
710
711 sigchain_push(SIGPIPE, SIG_IGN);
712
713 assert(strlen(filter_type) < LARGE_PACKET_DATA_MAX - strlen("command=\n"));
714 err = packet_write_fmt_gently(process->in, "command=%s\n", filter_type);
715 if (err)
716 goto done;
717
718 err = strlen(path) > LARGE_PACKET_DATA_MAX - strlen("pathname=\n");
719 if (err) {
720 error("path name too long for external filter");
721 goto done;
722 }
723
724 err = packet_write_fmt_gently(process->in, "pathname=%s\n", path);
725 if (err)
726 goto done;
727
728 err = packet_flush_gently(process->in);
729 if (err)
730 goto done;
731
732 if (fd >= 0)
733 err = write_packetized_from_fd(fd, process->in);
734 else
735 err = write_packetized_from_buf(src, len, process->in);
736 if (err)
737 goto done;
738
739 read_multi_file_filter_status(process->out, &filter_status);
740 err = strcmp(filter_status.buf, "success");
741 if (err)
742 goto done;
743
744 err = read_packetized_to_strbuf(process->out, &nbuf) < 0;
745 if (err)
746 goto done;
747
748 read_multi_file_filter_status(process->out, &filter_status);
749 err = strcmp(filter_status.buf, "success");
750
751 done:
752 sigchain_pop(SIGPIPE);
753
754 if (err || errno == EPIPE) {
755 if (!strcmp(filter_status.buf, "error")) {
756 /* The filter signaled a problem with the file. */
757 } else if (!strcmp(filter_status.buf, "abort")) {
758 /*
759 * The filter signaled a permanent problem. Don't try to filter
760 * files with the same command for the lifetime of the current
761 * Git process.
762 */
763 entry->supported_capabilities &= ~wanted_capability;
764 } else {
765 /*
766 * Something went wrong with the protocol filter.
767 * Force shutdown and restart if another blob requires filtering.
768 */
769 error("external filter '%s' failed", cmd);
770 kill_multi_file_filter(&cmd_process_map, entry);
771 }
772 } else {
773 strbuf_swap(dst, &nbuf);
774 }
775 strbuf_release(&nbuf);
776 return !err;
777 }
778
779 static struct convert_driver {
780 const char *name;
781 struct convert_driver *next;
782 const char *smudge;
783 const char *clean;
784 const char *process;
785 int required;
786 } *user_convert, **user_convert_tail;
787
788 static int apply_filter(const char *path, const char *src, size_t len,
789 int fd, struct strbuf *dst, struct convert_driver *drv,
790 const unsigned int wanted_capability)
791 {
792 const char *cmd = NULL;
793
794 if (!drv)
795 return 0;
796
797 if (!dst)
798 return 1;
799
800 if ((CAP_CLEAN & wanted_capability) && !drv->process && drv->clean)
801 cmd = drv->clean;
802 else if ((CAP_SMUDGE & wanted_capability) && !drv->process && drv->smudge)
803 cmd = drv->smudge;
804
805 if (cmd && *cmd)
806 return apply_single_file_filter(path, src, len, fd, dst, cmd);
807 else if (drv->process && *drv->process)
808 return apply_multi_file_filter(path, src, len, fd, dst, drv->process, wanted_capability);
809
810 return 0;
811 }
812
813 static int read_convert_config(const char *var, const char *value, void *cb)
814 {
815 const char *key, *name;
816 int namelen;
817 struct convert_driver *drv;
818
819 /*
820 * External conversion drivers are configured using
821 * "filter.<name>.variable".
822 */
823 if (parse_config_key(var, "filter", &name, &namelen, &key) < 0 || !name)
824 return 0;
825 for (drv = user_convert; drv; drv = drv->next)
826 if (!strncmp(drv->name, name, namelen) && !drv->name[namelen])
827 break;
828 if (!drv) {
829 drv = xcalloc(1, sizeof(struct convert_driver));
830 drv->name = xmemdupz(name, namelen);
831 *user_convert_tail = drv;
832 user_convert_tail = &(drv->next);
833 }
834
835 /*
836 * filter.<name>.smudge and filter.<name>.clean specifies
837 * the command line:
838 *
839 * command-line
840 *
841 * The command-line will not be interpolated in any way.
842 */
843
844 if (!strcmp("smudge", key))
845 return git_config_string(&drv->smudge, var, value);
846
847 if (!strcmp("clean", key))
848 return git_config_string(&drv->clean, var, value);
849
850 if (!strcmp("process", key))
851 return git_config_string(&drv->process, var, value);
852
853 if (!strcmp("required", key)) {
854 drv->required = git_config_bool(var, value);
855 return 0;
856 }
857
858 return 0;
859 }
860
861 static int count_ident(const char *cp, unsigned long size)
862 {
863 /*
864 * "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$"
865 */
866 int cnt = 0;
867 char ch;
868
869 while (size) {
870 ch = *cp++;
871 size--;
872 if (ch != '$')
873 continue;
874 if (size < 3)
875 break;
876 if (memcmp("Id", cp, 2))
877 continue;
878 ch = cp[2];
879 cp += 3;
880 size -= 3;
881 if (ch == '$')
882 cnt++; /* $Id$ */
883 if (ch != ':')
884 continue;
885
886 /*
887 * "$Id: ... "; scan up to the closing dollar sign and discard.
888 */
889 while (size) {
890 ch = *cp++;
891 size--;
892 if (ch == '$') {
893 cnt++;
894 break;
895 }
896 if (ch == '\n')
897 break;
898 }
899 }
900 return cnt;
901 }
902
903 static int ident_to_git(const char *path, const char *src, size_t len,
904 struct strbuf *buf, int ident)
905 {
906 char *dst, *dollar;
907
908 if (!ident || (src && !count_ident(src, len)))
909 return 0;
910
911 if (!buf)
912 return 1;
913
914 /* only grow if not in place */
915 if (strbuf_avail(buf) + buf->len < len)
916 strbuf_grow(buf, len - buf->len);
917 dst = buf->buf;
918 for (;;) {
919 dollar = memchr(src, '$', len);
920 if (!dollar)
921 break;
922 memmove(dst, src, dollar + 1 - src);
923 dst += dollar + 1 - src;
924 len -= dollar + 1 - src;
925 src = dollar + 1;
926
927 if (len > 3 && !memcmp(src, "Id:", 3)) {
928 dollar = memchr(src + 3, '$', len - 3);
929 if (!dollar)
930 break;
931 if (memchr(src + 3, '\n', dollar - src - 3)) {
932 /* Line break before the next dollar. */
933 continue;
934 }
935
936 memcpy(dst, "Id$", 3);
937 dst += 3;
938 len -= dollar + 1 - src;
939 src = dollar + 1;
940 }
941 }
942 memmove(dst, src, len);
943 strbuf_setlen(buf, dst + len - buf->buf);
944 return 1;
945 }
946
947 static int ident_to_worktree(const char *path, const char *src, size_t len,
948 struct strbuf *buf, int ident)
949 {
950 unsigned char sha1[20];
951 char *to_free = NULL, *dollar, *spc;
952 int cnt;
953
954 if (!ident)
955 return 0;
956
957 cnt = count_ident(src, len);
958 if (!cnt)
959 return 0;
960
961 /* are we "faking" in place editing ? */
962 if (src == buf->buf)
963 to_free = strbuf_detach(buf, NULL);
964 hash_sha1_file(src, len, "blob", sha1);
965
966 strbuf_grow(buf, len + cnt * 43);
967 for (;;) {
968 /* step 1: run to the next '$' */
969 dollar = memchr(src, '$', len);
970 if (!dollar)
971 break;
972 strbuf_add(buf, src, dollar + 1 - src);
973 len -= dollar + 1 - src;
974 src = dollar + 1;
975
976 /* step 2: does it looks like a bit like Id:xxx$ or Id$ ? */
977 if (len < 3 || memcmp("Id", src, 2))
978 continue;
979
980 /* step 3: skip over Id$ or Id:xxxxx$ */
981 if (src[2] == '$') {
982 src += 3;
983 len -= 3;
984 } else if (src[2] == ':') {
985 /*
986 * It's possible that an expanded Id has crept its way into the
987 * repository, we cope with that by stripping the expansion out.
988 * This is probably not a good idea, since it will cause changes
989 * on checkout, which won't go away by stash, but let's keep it
990 * for git-style ids.
991 */
992 dollar = memchr(src + 3, '$', len - 3);
993 if (!dollar) {
994 /* incomplete keyword, no more '$', so just quit the loop */
995 break;
996 }
997
998 if (memchr(src + 3, '\n', dollar - src - 3)) {
999 /* Line break before the next dollar. */
1000 continue;
1001 }
1002
1003 spc = memchr(src + 4, ' ', dollar - src - 4);
1004 if (spc && spc < dollar-1) {
1005 /* There are spaces in unexpected places.
1006 * This is probably an id from some other
1007 * versioning system. Keep it for now.
1008 */
1009 continue;
1010 }
1011
1012 len -= dollar + 1 - src;
1013 src = dollar + 1;
1014 } else {
1015 /* it wasn't a "Id$" or "Id:xxxx$" */
1016 continue;
1017 }
1018
1019 /* step 4: substitute */
1020 strbuf_addstr(buf, "Id: ");
1021 strbuf_add(buf, sha1_to_hex(sha1), 40);
1022 strbuf_addstr(buf, " $");
1023 }
1024 strbuf_add(buf, src, len);
1025
1026 free(to_free);
1027 return 1;
1028 }
1029
1030 static enum crlf_action git_path_check_crlf(struct git_attr_check *check)
1031 {
1032 const char *value = check->value;
1033
1034 if (ATTR_TRUE(value))
1035 return CRLF_TEXT;
1036 else if (ATTR_FALSE(value))
1037 return CRLF_BINARY;
1038 else if (ATTR_UNSET(value))
1039 ;
1040 else if (!strcmp(value, "input"))
1041 return CRLF_TEXT_INPUT;
1042 else if (!strcmp(value, "auto"))
1043 return CRLF_AUTO;
1044 return CRLF_UNDEFINED;
1045 }
1046
1047 static enum eol git_path_check_eol(struct git_attr_check *check)
1048 {
1049 const char *value = check->value;
1050
1051 if (ATTR_UNSET(value))
1052 ;
1053 else if (!strcmp(value, "lf"))
1054 return EOL_LF;
1055 else if (!strcmp(value, "crlf"))
1056 return EOL_CRLF;
1057 return EOL_UNSET;
1058 }
1059
1060 static struct convert_driver *git_path_check_convert(struct git_attr_check *check)
1061 {
1062 const char *value = check->value;
1063 struct convert_driver *drv;
1064
1065 if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
1066 return NULL;
1067 for (drv = user_convert; drv; drv = drv->next)
1068 if (!strcmp(value, drv->name))
1069 return drv;
1070 return NULL;
1071 }
1072
1073 static int git_path_check_ident(struct git_attr_check *check)
1074 {
1075 const char *value = check->value;
1076
1077 return !!ATTR_TRUE(value);
1078 }
1079
1080 struct conv_attrs {
1081 struct convert_driver *drv;
1082 enum crlf_action attr_action; /* What attr says */
1083 enum crlf_action crlf_action; /* When no attr is set, use core.autocrlf */
1084 int ident;
1085 };
1086
1087 static const char *conv_attr_name[] = {
1088 "crlf", "ident", "filter", "eol", "text",
1089 };
1090 #define NUM_CONV_ATTRS ARRAY_SIZE(conv_attr_name)
1091
1092 static void convert_attrs(struct conv_attrs *ca, const char *path)
1093 {
1094 int i;
1095 static struct git_attr_check ccheck[NUM_CONV_ATTRS];
1096
1097 if (!ccheck[0].attr) {
1098 for (i = 0; i < NUM_CONV_ATTRS; i++)
1099 ccheck[i].attr = git_attr(conv_attr_name[i]);
1100 user_convert_tail = &user_convert;
1101 git_config(read_convert_config, NULL);
1102 }
1103
1104 if (!git_check_attr(path, NUM_CONV_ATTRS, ccheck)) {
1105 ca->crlf_action = git_path_check_crlf(ccheck + 4);
1106 if (ca->crlf_action == CRLF_UNDEFINED)
1107 ca->crlf_action = git_path_check_crlf(ccheck + 0);
1108 ca->attr_action = ca->crlf_action;
1109 ca->ident = git_path_check_ident(ccheck + 1);
1110 ca->drv = git_path_check_convert(ccheck + 2);
1111 if (ca->crlf_action != CRLF_BINARY) {
1112 enum eol eol_attr = git_path_check_eol(ccheck + 3);
1113 if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_LF)
1114 ca->crlf_action = CRLF_AUTO_INPUT;
1115 else if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_CRLF)
1116 ca->crlf_action = CRLF_AUTO_CRLF;
1117 else if (eol_attr == EOL_LF)
1118 ca->crlf_action = CRLF_TEXT_INPUT;
1119 else if (eol_attr == EOL_CRLF)
1120 ca->crlf_action = CRLF_TEXT_CRLF;
1121 }
1122 ca->attr_action = ca->crlf_action;
1123 } else {
1124 ca->drv = NULL;
1125 ca->crlf_action = CRLF_UNDEFINED;
1126 ca->ident = 0;
1127 }
1128 if (ca->crlf_action == CRLF_TEXT)
1129 ca->crlf_action = text_eol_is_crlf() ? CRLF_TEXT_CRLF : CRLF_TEXT_INPUT;
1130 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_FALSE)
1131 ca->crlf_action = CRLF_BINARY;
1132 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_TRUE)
1133 ca->crlf_action = CRLF_AUTO_CRLF;
1134 if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_INPUT)
1135 ca->crlf_action = CRLF_AUTO_INPUT;
1136 }
1137
1138 int would_convert_to_git_filter_fd(const char *path)
1139 {
1140 struct conv_attrs ca;
1141
1142 convert_attrs(&ca, path);
1143 if (!ca.drv)
1144 return 0;
1145
1146 /*
1147 * Apply a filter to an fd only if the filter is required to succeed.
1148 * We must die if the filter fails, because the original data before
1149 * filtering is not available.
1150 */
1151 if (!ca.drv->required)
1152 return 0;
1153
1154 return apply_filter(path, NULL, 0, -1, NULL, ca.drv, CAP_CLEAN);
1155 }
1156
1157 const char *get_convert_attr_ascii(const char *path)
1158 {
1159 struct conv_attrs ca;
1160
1161 convert_attrs(&ca, path);
1162 switch (ca.attr_action) {
1163 case CRLF_UNDEFINED:
1164 return "";
1165 case CRLF_BINARY:
1166 return "-text";
1167 case CRLF_TEXT:
1168 return "text";
1169 case CRLF_TEXT_INPUT:
1170 return "text eol=lf";
1171 case CRLF_TEXT_CRLF:
1172 return "text eol=crlf";
1173 case CRLF_AUTO:
1174 return "text=auto";
1175 case CRLF_AUTO_CRLF:
1176 return "text=auto eol=crlf";
1177 case CRLF_AUTO_INPUT:
1178 return "text=auto eol=lf";
1179 }
1180 return "";
1181 }
1182
1183 int convert_to_git(const char *path, const char *src, size_t len,
1184 struct strbuf *dst, enum safe_crlf checksafe)
1185 {
1186 int ret = 0;
1187 struct conv_attrs ca;
1188
1189 convert_attrs(&ca, path);
1190
1191 ret |= apply_filter(path, src, len, -1, dst, ca.drv, CAP_CLEAN);
1192 if (!ret && ca.drv && ca.drv->required)
1193 die("%s: clean filter '%s' failed", path, ca.drv->name);
1194
1195 if (ret && dst) {
1196 src = dst->buf;
1197 len = dst->len;
1198 }
1199 ret |= crlf_to_git(path, src, len, dst, ca.crlf_action, checksafe);
1200 if (ret && dst) {
1201 src = dst->buf;
1202 len = dst->len;
1203 }
1204 return ret | ident_to_git(path, src, len, dst, ca.ident);
1205 }
1206
1207 void convert_to_git_filter_fd(const char *path, int fd, struct strbuf *dst,
1208 enum safe_crlf checksafe)
1209 {
1210 struct conv_attrs ca;
1211 convert_attrs(&ca, path);
1212
1213 assert(ca.drv);
1214 assert(ca.drv->clean || ca.drv->process);
1215
1216 if (!apply_filter(path, NULL, 0, fd, dst, ca.drv, CAP_CLEAN))
1217 die("%s: clean filter '%s' failed", path, ca.drv->name);
1218
1219 crlf_to_git(path, dst->buf, dst->len, dst, ca.crlf_action, checksafe);
1220 ident_to_git(path, dst->buf, dst->len, dst, ca.ident);
1221 }
1222
1223 static int convert_to_working_tree_internal(const char *path, const char *src,
1224 size_t len, struct strbuf *dst,
1225 int normalizing)
1226 {
1227 int ret = 0, ret_filter = 0;
1228 struct conv_attrs ca;
1229
1230 convert_attrs(&ca, path);
1231
1232 ret |= ident_to_worktree(path, src, len, dst, ca.ident);
1233 if (ret) {
1234 src = dst->buf;
1235 len = dst->len;
1236 }
1237 /*
1238 * CRLF conversion can be skipped if normalizing, unless there
1239 * is a smudge or process filter (even if the process filter doesn't
1240 * support smudge). The filters might expect CRLFs.
1241 */
1242 if ((ca.drv && (ca.drv->smudge || ca.drv->process)) || !normalizing) {
1243 ret |= crlf_to_worktree(path, src, len, dst, ca.crlf_action);
1244 if (ret) {
1245 src = dst->buf;
1246 len = dst->len;
1247 }
1248 }
1249
1250 ret_filter = apply_filter(path, src, len, -1, dst, ca.drv, CAP_SMUDGE);
1251 if (!ret_filter && ca.drv && ca.drv->required)
1252 die("%s: smudge filter %s failed", path, ca.drv->name);
1253
1254 return ret | ret_filter;
1255 }
1256
1257 int convert_to_working_tree(const char *path, const char *src, size_t len, struct strbuf *dst)
1258 {
1259 return convert_to_working_tree_internal(path, src, len, dst, 0);
1260 }
1261
1262 int renormalize_buffer(const char *path, const char *src, size_t len, struct strbuf *dst)
1263 {
1264 int ret = convert_to_working_tree_internal(path, src, len, dst, 1);
1265 if (ret) {
1266 src = dst->buf;
1267 len = dst->len;
1268 }
1269 return ret | convert_to_git(path, src, len, dst, SAFE_CRLF_RENORMALIZE);
1270 }
1271
1272 /*****************************************************************
1273 *
1274 * Streaming conversion support
1275 *
1276 *****************************************************************/
1277
1278 typedef int (*filter_fn)(struct stream_filter *,
1279 const char *input, size_t *isize_p,
1280 char *output, size_t *osize_p);
1281 typedef void (*free_fn)(struct stream_filter *);
1282
1283 struct stream_filter_vtbl {
1284 filter_fn filter;
1285 free_fn free;
1286 };
1287
1288 struct stream_filter {
1289 struct stream_filter_vtbl *vtbl;
1290 };
1291
1292 static int null_filter_fn(struct stream_filter *filter,
1293 const char *input, size_t *isize_p,
1294 char *output, size_t *osize_p)
1295 {
1296 size_t count;
1297
1298 if (!input)
1299 return 0; /* we do not keep any states */
1300 count = *isize_p;
1301 if (*osize_p < count)
1302 count = *osize_p;
1303 if (count) {
1304 memmove(output, input, count);
1305 *isize_p -= count;
1306 *osize_p -= count;
1307 }
1308 return 0;
1309 }
1310
1311 static void null_free_fn(struct stream_filter *filter)
1312 {
1313 ; /* nothing -- null instances are shared */
1314 }
1315
1316 static struct stream_filter_vtbl null_vtbl = {
1317 null_filter_fn,
1318 null_free_fn,
1319 };
1320
1321 static struct stream_filter null_filter_singleton = {
1322 &null_vtbl,
1323 };
1324
1325 int is_null_stream_filter(struct stream_filter *filter)
1326 {
1327 return filter == &null_filter_singleton;
1328 }
1329
1330
1331 /*
1332 * LF-to-CRLF filter
1333 */
1334
1335 struct lf_to_crlf_filter {
1336 struct stream_filter filter;
1337 unsigned has_held:1;
1338 char held;
1339 };
1340
1341 static int lf_to_crlf_filter_fn(struct stream_filter *filter,
1342 const char *input, size_t *isize_p,
1343 char *output, size_t *osize_p)
1344 {
1345 size_t count, o = 0;
1346 struct lf_to_crlf_filter *lf_to_crlf = (struct lf_to_crlf_filter *)filter;
1347
1348 /*
1349 * We may be holding onto the CR to see if it is followed by a
1350 * LF, in which case we would need to go to the main loop.
1351 * Otherwise, just emit it to the output stream.
1352 */
1353 if (lf_to_crlf->has_held && (lf_to_crlf->held != '\r' || !input)) {
1354 output[o++] = lf_to_crlf->held;
1355 lf_to_crlf->has_held = 0;
1356 }
1357
1358 /* We are told to drain */
1359 if (!input) {
1360 *osize_p -= o;
1361 return 0;
1362 }
1363
1364 count = *isize_p;
1365 if (count || lf_to_crlf->has_held) {
1366 size_t i;
1367 int was_cr = 0;
1368
1369 if (lf_to_crlf->has_held) {
1370 was_cr = 1;
1371 lf_to_crlf->has_held = 0;
1372 }
1373
1374 for (i = 0; o < *osize_p && i < count; i++) {
1375 char ch = input[i];
1376
1377 if (ch == '\n') {
1378 output[o++] = '\r';
1379 } else if (was_cr) {
1380 /*
1381 * Previous round saw CR and it is not followed
1382 * by a LF; emit the CR before processing the
1383 * current character.
1384 */
1385 output[o++] = '\r';
1386 }
1387
1388 /*
1389 * We may have consumed the last output slot,
1390 * in which case we need to break out of this
1391 * loop; hold the current character before
1392 * returning.
1393 */
1394 if (*osize_p <= o) {
1395 lf_to_crlf->has_held = 1;
1396 lf_to_crlf->held = ch;
1397 continue; /* break but increment i */
1398 }
1399
1400 if (ch == '\r') {
1401 was_cr = 1;
1402 continue;
1403 }
1404
1405 was_cr = 0;
1406 output[o++] = ch;
1407 }
1408
1409 *osize_p -= o;
1410 *isize_p -= i;
1411
1412 if (!lf_to_crlf->has_held && was_cr) {
1413 lf_to_crlf->has_held = 1;
1414 lf_to_crlf->held = '\r';
1415 }
1416 }
1417 return 0;
1418 }
1419
1420 static void lf_to_crlf_free_fn(struct stream_filter *filter)
1421 {
1422 free(filter);
1423 }
1424
1425 static struct stream_filter_vtbl lf_to_crlf_vtbl = {
1426 lf_to_crlf_filter_fn,
1427 lf_to_crlf_free_fn,
1428 };
1429
1430 static struct stream_filter *lf_to_crlf_filter(void)
1431 {
1432 struct lf_to_crlf_filter *lf_to_crlf = xcalloc(1, sizeof(*lf_to_crlf));
1433
1434 lf_to_crlf->filter.vtbl = &lf_to_crlf_vtbl;
1435 return (struct stream_filter *)lf_to_crlf;
1436 }
1437
1438 /*
1439 * Cascade filter
1440 */
1441 #define FILTER_BUFFER 1024
1442 struct cascade_filter {
1443 struct stream_filter filter;
1444 struct stream_filter *one;
1445 struct stream_filter *two;
1446 char buf[FILTER_BUFFER];
1447 int end, ptr;
1448 };
1449
1450 static int cascade_filter_fn(struct stream_filter *filter,
1451 const char *input, size_t *isize_p,
1452 char *output, size_t *osize_p)
1453 {
1454 struct cascade_filter *cas = (struct cascade_filter *) filter;
1455 size_t filled = 0;
1456 size_t sz = *osize_p;
1457 size_t to_feed, remaining;
1458
1459 /*
1460 * input -- (one) --> buf -- (two) --> output
1461 */
1462 while (filled < sz) {
1463 remaining = sz - filled;
1464
1465 /* do we already have something to feed two with? */
1466 if (cas->ptr < cas->end) {
1467 to_feed = cas->end - cas->ptr;
1468 if (stream_filter(cas->two,
1469 cas->buf + cas->ptr, &to_feed,
1470 output + filled, &remaining))
1471 return -1;
1472 cas->ptr += (cas->end - cas->ptr) - to_feed;
1473 filled = sz - remaining;
1474 continue;
1475 }
1476
1477 /* feed one from upstream and have it emit into our buffer */
1478 to_feed = input ? *isize_p : 0;
1479 if (input && !to_feed)
1480 break;
1481 remaining = sizeof(cas->buf);
1482 if (stream_filter(cas->one,
1483 input, &to_feed,
1484 cas->buf, &remaining))
1485 return -1;
1486 cas->end = sizeof(cas->buf) - remaining;
1487 cas->ptr = 0;
1488 if (input) {
1489 size_t fed = *isize_p - to_feed;
1490 *isize_p -= fed;
1491 input += fed;
1492 }
1493
1494 /* do we know that we drained one completely? */
1495 if (input || cas->end)
1496 continue;
1497
1498 /* tell two to drain; we have nothing more to give it */
1499 to_feed = 0;
1500 remaining = sz - filled;
1501 if (stream_filter(cas->two,
1502 NULL, &to_feed,
1503 output + filled, &remaining))
1504 return -1;
1505 if (remaining == (sz - filled))
1506 break; /* completely drained two */
1507 filled = sz - remaining;
1508 }
1509 *osize_p -= filled;
1510 return 0;
1511 }
1512
1513 static void cascade_free_fn(struct stream_filter *filter)
1514 {
1515 struct cascade_filter *cas = (struct cascade_filter *)filter;
1516 free_stream_filter(cas->one);
1517 free_stream_filter(cas->two);
1518 free(filter);
1519 }
1520
1521 static struct stream_filter_vtbl cascade_vtbl = {
1522 cascade_filter_fn,
1523 cascade_free_fn,
1524 };
1525
1526 static struct stream_filter *cascade_filter(struct stream_filter *one,
1527 struct stream_filter *two)
1528 {
1529 struct cascade_filter *cascade;
1530
1531 if (!one || is_null_stream_filter(one))
1532 return two;
1533 if (!two || is_null_stream_filter(two))
1534 return one;
1535
1536 cascade = xmalloc(sizeof(*cascade));
1537 cascade->one = one;
1538 cascade->two = two;
1539 cascade->end = cascade->ptr = 0;
1540 cascade->filter.vtbl = &cascade_vtbl;
1541 return (struct stream_filter *)cascade;
1542 }
1543
1544 /*
1545 * ident filter
1546 */
1547 #define IDENT_DRAINING (-1)
1548 #define IDENT_SKIPPING (-2)
1549 struct ident_filter {
1550 struct stream_filter filter;
1551 struct strbuf left;
1552 int state;
1553 char ident[45]; /* ": x40 $" */
1554 };
1555
1556 static int is_foreign_ident(const char *str)
1557 {
1558 int i;
1559
1560 if (!skip_prefix(str, "$Id: ", &str))
1561 return 0;
1562 for (i = 0; str[i]; i++) {
1563 if (isspace(str[i]) && str[i+1] != '$')
1564 return 1;
1565 }
1566 return 0;
1567 }
1568
1569 static void ident_drain(struct ident_filter *ident, char **output_p, size_t *osize_p)
1570 {
1571 size_t to_drain = ident->left.len;
1572
1573 if (*osize_p < to_drain)
1574 to_drain = *osize_p;
1575 if (to_drain) {
1576 memcpy(*output_p, ident->left.buf, to_drain);
1577 strbuf_remove(&ident->left, 0, to_drain);
1578 *output_p += to_drain;
1579 *osize_p -= to_drain;
1580 }
1581 if (!ident->left.len)
1582 ident->state = 0;
1583 }
1584
1585 static int ident_filter_fn(struct stream_filter *filter,
1586 const char *input, size_t *isize_p,
1587 char *output, size_t *osize_p)
1588 {
1589 struct ident_filter *ident = (struct ident_filter *)filter;
1590 static const char head[] = "$Id";
1591
1592 if (!input) {
1593 /* drain upon eof */
1594 switch (ident->state) {
1595 default:
1596 strbuf_add(&ident->left, head, ident->state);
1597 case IDENT_SKIPPING:
1598 /* fallthru */
1599 case IDENT_DRAINING:
1600 ident_drain(ident, &output, osize_p);
1601 }
1602 return 0;
1603 }
1604
1605 while (*isize_p || (ident->state == IDENT_DRAINING)) {
1606 int ch;
1607
1608 if (ident->state == IDENT_DRAINING) {
1609 ident_drain(ident, &output, osize_p);
1610 if (!*osize_p)
1611 break;
1612 continue;
1613 }
1614
1615 ch = *(input++);
1616 (*isize_p)--;
1617
1618 if (ident->state == IDENT_SKIPPING) {
1619 /*
1620 * Skipping until '$' or LF, but keeping them
1621 * in case it is a foreign ident.
1622 */
1623 strbuf_addch(&ident->left, ch);
1624 if (ch != '\n' && ch != '$')
1625 continue;
1626 if (ch == '$' && !is_foreign_ident(ident->left.buf)) {
1627 strbuf_setlen(&ident->left, sizeof(head) - 1);
1628 strbuf_addstr(&ident->left, ident->ident);
1629 }
1630 ident->state = IDENT_DRAINING;
1631 continue;
1632 }
1633
1634 if (ident->state < sizeof(head) &&
1635 head[ident->state] == ch) {
1636 ident->state++;
1637 continue;
1638 }
1639
1640 if (ident->state)
1641 strbuf_add(&ident->left, head, ident->state);
1642 if (ident->state == sizeof(head) - 1) {
1643 if (ch != ':' && ch != '$') {
1644 strbuf_addch(&ident->left, ch);
1645 ident->state = 0;
1646 continue;
1647 }
1648
1649 if (ch == ':') {
1650 strbuf_addch(&ident->left, ch);
1651 ident->state = IDENT_SKIPPING;
1652 } else {
1653 strbuf_addstr(&ident->left, ident->ident);
1654 ident->state = IDENT_DRAINING;
1655 }
1656 continue;
1657 }
1658
1659 strbuf_addch(&ident->left, ch);
1660 ident->state = IDENT_DRAINING;
1661 }
1662 return 0;
1663 }
1664
1665 static void ident_free_fn(struct stream_filter *filter)
1666 {
1667 struct ident_filter *ident = (struct ident_filter *)filter;
1668 strbuf_release(&ident->left);
1669 free(filter);
1670 }
1671
1672 static struct stream_filter_vtbl ident_vtbl = {
1673 ident_filter_fn,
1674 ident_free_fn,
1675 };
1676
1677 static struct stream_filter *ident_filter(const unsigned char *sha1)
1678 {
1679 struct ident_filter *ident = xmalloc(sizeof(*ident));
1680
1681 xsnprintf(ident->ident, sizeof(ident->ident),
1682 ": %s $", sha1_to_hex(sha1));
1683 strbuf_init(&ident->left, 0);
1684 ident->filter.vtbl = &ident_vtbl;
1685 ident->state = 0;
1686 return (struct stream_filter *)ident;
1687 }
1688
1689 /*
1690 * Return an appropriately constructed filter for the path, or NULL if
1691 * the contents cannot be filtered without reading the whole thing
1692 * in-core.
1693 *
1694 * Note that you would be crazy to set CRLF, smuge/clean or ident to a
1695 * large binary blob you would want us not to slurp into the memory!
1696 */
1697 struct stream_filter *get_stream_filter(const char *path, const unsigned char *sha1)
1698 {
1699 struct conv_attrs ca;
1700 struct stream_filter *filter = NULL;
1701
1702 convert_attrs(&ca, path);
1703 if (ca.drv && (ca.drv->process || ca.drv->smudge || ca.drv->clean))
1704 return NULL;
1705
1706 if (ca.crlf_action == CRLF_AUTO || ca.crlf_action == CRLF_AUTO_CRLF)
1707 return NULL;
1708
1709 if (ca.ident)
1710 filter = ident_filter(sha1);
1711
1712 if (output_eol(ca.crlf_action) == EOL_CRLF)
1713 filter = cascade_filter(filter, lf_to_crlf_filter());
1714 else
1715 filter = cascade_filter(filter, &null_filter_singleton);
1716
1717 return filter;
1718 }
1719
1720 void free_stream_filter(struct stream_filter *filter)
1721 {
1722 filter->vtbl->free(filter);
1723 }
1724
1725 int stream_filter(struct stream_filter *filter,
1726 const char *input, size_t *isize_p,
1727 char *output, size_t *osize_p)
1728 {
1729 return filter->vtbl->filter(filter, input, isize_p, output, osize_p);
1730 }
Eric Wong's git-svn branch