treat any file with NUL as binary
[git/git.git] / convert.c
1 #include "cache.h"
2 #include "attr.h"
3 #include "run-command.h"
4
5 /*
6 * convert.c - convert a file when checking it out and checking it in.
7 *
8 * This should use the pathname to decide on whether it wants to do some
9 * more interesting conversions (automatic gzip/unzip, general format
10 * conversions etc etc), but by default it just does automatic CRLF<->LF
11 * translation when the "auto_crlf" option is set.
12 */
13
14 #define CRLF_GUESS (-1)
15 #define CRLF_BINARY 0
16 #define CRLF_TEXT 1
17 #define CRLF_INPUT 2
18
19 struct text_stat {
20 /* NUL, CR, LF and CRLF counts */
21 unsigned nul, cr, lf, crlf;
22
23 /* These are just approximations! */
24 unsigned printable, nonprintable;
25 };
26
27 static void gather_stats(const char *buf, unsigned long size, struct text_stat *stats)
28 {
29 unsigned long i;
30
31 memset(stats, 0, sizeof(*stats));
32
33 for (i = 0; i < size; i++) {
34 unsigned char c = buf[i];
35 if (c == '\r') {
36 stats->cr++;
37 if (i+1 < size && buf[i+1] == '\n')
38 stats->crlf++;
39 continue;
40 }
41 if (c == '\n') {
42 stats->lf++;
43 continue;
44 }
45 if (c == 127)
46 /* DEL */
47 stats->nonprintable++;
48 else if (c < 32) {
49 switch (c) {
50 /* BS, HT, ESC and FF */
51 case '\b': case '\t': case '\033': case '\014':
52 stats->printable++;
53 break;
54 case 0:
55 stats->nul++;
56 /* fall through */
57 default:
58 stats->nonprintable++;
59 }
60 }
61 else
62 stats->printable++;
63 }
64 }
65
66 /*
67 * The same heuristics as diff.c::mmfile_is_binary()
68 */
69 static int is_binary(unsigned long size, struct text_stat *stats)
70 {
71
72 if (stats->nul)
73 return 1;
74 if ((stats->printable >> 7) < stats->nonprintable)
75 return 1;
76 /*
77 * Other heuristics? Average line length might be relevant,
78 * as might LF vs CR vs CRLF counts..
79 *
80 * NOTE! It might be normal to have a low ratio of CRLF to LF
81 * (somebody starts with a LF-only file and edits it with an editor
82 * that adds CRLF only to lines that are added..). But do we
83 * want to support CR-only? Probably not.
84 */
85 return 0;
86 }
87
88 static int crlf_to_git(const char *path, const char *src, size_t len,
89 struct strbuf *buf, int action)
90 {
91 struct text_stat stats;
92 char *dst;
93
94 if ((action == CRLF_BINARY) || !auto_crlf || !len)
95 return 0;
96
97 gather_stats(src, len, &stats);
98 /* No CR? Nothing to convert, regardless. */
99 if (!stats.cr)
100 return 0;
101
102 if (action == CRLF_GUESS) {
103 /*
104 * We're currently not going to even try to convert stuff
105 * that has bare CR characters. Does anybody do that crazy
106 * stuff?
107 */
108 if (stats.cr != stats.crlf)
109 return 0;
110
111 /*
112 * And add some heuristics for binary vs text, of course...
113 */
114 if (is_binary(len, &stats))
115 return 0;
116 }
117
118 /* only grow if not in place */
119 if (strbuf_avail(buf) + buf->len < len)
120 strbuf_grow(buf, len - buf->len);
121 dst = buf->buf;
122 if (action == CRLF_GUESS) {
123 /*
124 * If we guessed, we already know we rejected a file with
125 * lone CR, and we can strip a CR without looking at what
126 * follow it.
127 */
128 do {
129 unsigned char c = *src++;
130 if (c != '\r')
131 *dst++ = c;
132 } while (--len);
133 } else {
134 do {
135 unsigned char c = *src++;
136 if (! (c == '\r' && (1 < len && *src == '\n')))
137 *dst++ = c;
138 } while (--len);
139 }
140 strbuf_setlen(buf, dst - buf->buf);
141 return 1;
142 }
143
144 static int crlf_to_worktree(const char *path, const char *src, size_t len,
145 struct strbuf *buf, int action)
146 {
147 char *to_free = NULL;
148 struct text_stat stats;
149
150 if ((action == CRLF_BINARY) || (action == CRLF_INPUT) ||
151 auto_crlf <= 0)
152 return 0;
153
154 if (!len)
155 return 0;
156
157 gather_stats(src, len, &stats);
158
159 /* No LF? Nothing to convert, regardless. */
160 if (!stats.lf)
161 return 0;
162
163 /* Was it already in CRLF format? */
164 if (stats.lf == stats.crlf)
165 return 0;
166
167 if (action == CRLF_GUESS) {
168 /* If we have any bare CR characters, we're not going to touch it */
169 if (stats.cr != stats.crlf)
170 return 0;
171
172 if (is_binary(len, &stats))
173 return 0;
174 }
175
176 /* are we "faking" in place editing ? */
177 if (src == buf->buf)
178 to_free = strbuf_detach(buf, NULL);
179
180 strbuf_grow(buf, len + stats.lf - stats.crlf);
181 for (;;) {
182 const char *nl = memchr(src, '\n', len);
183 if (!nl)
184 break;
185 if (nl > src && nl[-1] == '\r') {
186 strbuf_add(buf, src, nl + 1 - src);
187 } else {
188 strbuf_add(buf, src, nl - src);
189 strbuf_addstr(buf, "\r\n");
190 }
191 len -= nl + 1 - src;
192 src = nl + 1;
193 }
194 strbuf_add(buf, src, len);
195
196 free(to_free);
197 return 1;
198 }
199
200 struct filter_params {
201 const char *src;
202 unsigned long size;
203 const char *cmd;
204 };
205
206 static int filter_buffer(int fd, void *data)
207 {
208 /*
209 * Spawn cmd and feed the buffer contents through its stdin.
210 */
211 struct child_process child_process;
212 struct filter_params *params = (struct filter_params *)data;
213 int write_err, status;
214 const char *argv[] = { "sh", "-c", params->cmd, NULL };
215
216 memset(&child_process, 0, sizeof(child_process));
217 child_process.argv = argv;
218 child_process.in = -1;
219 child_process.out = fd;
220
221 if (start_command(&child_process))
222 return error("cannot fork to run external filter %s", params->cmd);
223
224 write_err = (write_in_full(child_process.in, params->src, params->size) < 0);
225 if (close(child_process.in))
226 write_err = 1;
227 if (write_err)
228 error("cannot feed the input to external filter %s", params->cmd);
229
230 status = finish_command(&child_process);
231 if (status)
232 error("external filter %s failed %d", params->cmd, -status);
233 return (write_err || status);
234 }
235
236 static int apply_filter(const char *path, const char *src, size_t len,
237 struct strbuf *dst, const char *cmd)
238 {
239 /*
240 * Create a pipeline to have the command filter the buffer's
241 * contents.
242 *
243 * (child --> cmd) --> us
244 */
245 int ret = 1;
246 struct strbuf nbuf;
247 struct async async;
248 struct filter_params params;
249
250 if (!cmd)
251 return 0;
252
253 memset(&async, 0, sizeof(async));
254 async.proc = filter_buffer;
255 async.data = &params;
256 params.src = src;
257 params.size = len;
258 params.cmd = cmd;
259
260 fflush(NULL);
261 if (start_async(&async))
262 return 0; /* error was already reported */
263
264 strbuf_init(&nbuf, 0);
265 if (strbuf_read(&nbuf, async.out, len) < 0) {
266 error("read from external filter %s failed", cmd);
267 ret = 0;
268 }
269 if (close(async.out)) {
270 error("read from external filter %s failed", cmd);
271 ret = 0;
272 }
273 if (finish_async(&async)) {
274 error("external filter %s failed", cmd);
275 ret = 0;
276 }
277
278 if (ret) {
279 strbuf_swap(dst, &nbuf);
280 }
281 strbuf_release(&nbuf);
282 return ret;
283 }
284
285 static struct convert_driver {
286 const char *name;
287 struct convert_driver *next;
288 char *smudge;
289 char *clean;
290 } *user_convert, **user_convert_tail;
291
292 static int read_convert_config(const char *var, const char *value)
293 {
294 const char *ep, *name;
295 int namelen;
296 struct convert_driver *drv;
297
298 /*
299 * External conversion drivers are configured using
300 * "filter.<name>.variable".
301 */
302 if (prefixcmp(var, "filter.") || (ep = strrchr(var, '.')) == var + 6)
303 return 0;
304 name = var + 7;
305 namelen = ep - name;
306 for (drv = user_convert; drv; drv = drv->next)
307 if (!strncmp(drv->name, name, namelen) && !drv->name[namelen])
308 break;
309 if (!drv) {
310 drv = xcalloc(1, sizeof(struct convert_driver));
311 drv->name = xmemdupz(name, namelen);
312 *user_convert_tail = drv;
313 user_convert_tail = &(drv->next);
314 }
315
316 ep++;
317
318 /*
319 * filter.<name>.smudge and filter.<name>.clean specifies
320 * the command line:
321 *
322 * command-line
323 *
324 * The command-line will not be interpolated in any way.
325 */
326
327 if (!strcmp("smudge", ep)) {
328 if (!value)
329 return error("%s: lacks value", var);
330 drv->smudge = strdup(value);
331 return 0;
332 }
333
334 if (!strcmp("clean", ep)) {
335 if (!value)
336 return error("%s: lacks value", var);
337 drv->clean = strdup(value);
338 return 0;
339 }
340 return 0;
341 }
342
343 static void setup_convert_check(struct git_attr_check *check)
344 {
345 static struct git_attr *attr_crlf;
346 static struct git_attr *attr_ident;
347 static struct git_attr *attr_filter;
348
349 if (!attr_crlf) {
350 attr_crlf = git_attr("crlf", 4);
351 attr_ident = git_attr("ident", 5);
352 attr_filter = git_attr("filter", 6);
353 user_convert_tail = &user_convert;
354 git_config(read_convert_config);
355 }
356 check[0].attr = attr_crlf;
357 check[1].attr = attr_ident;
358 check[2].attr = attr_filter;
359 }
360
361 static int count_ident(const char *cp, unsigned long size)
362 {
363 /*
364 * "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$"
365 */
366 int cnt = 0;
367 char ch;
368
369 while (size) {
370 ch = *cp++;
371 size--;
372 if (ch != '$')
373 continue;
374 if (size < 3)
375 break;
376 if (memcmp("Id", cp, 2))
377 continue;
378 ch = cp[2];
379 cp += 3;
380 size -= 3;
381 if (ch == '$')
382 cnt++; /* $Id$ */
383 if (ch != ':')
384 continue;
385
386 /*
387 * "$Id: ... "; scan up to the closing dollar sign and discard.
388 */
389 while (size) {
390 ch = *cp++;
391 size--;
392 if (ch == '$') {
393 cnt++;
394 break;
395 }
396 }
397 }
398 return cnt;
399 }
400
401 static int ident_to_git(const char *path, const char *src, size_t len,
402 struct strbuf *buf, int ident)
403 {
404 char *dst, *dollar;
405
406 if (!ident || !count_ident(src, len))
407 return 0;
408
409 /* only grow if not in place */
410 if (strbuf_avail(buf) + buf->len < len)
411 strbuf_grow(buf, len - buf->len);
412 dst = buf->buf;
413 for (;;) {
414 dollar = memchr(src, '$', len);
415 if (!dollar)
416 break;
417 memcpy(dst, src, dollar + 1 - src);
418 dst += dollar + 1 - src;
419 len -= dollar + 1 - src;
420 src = dollar + 1;
421
422 if (len > 3 && !memcmp(src, "Id:", 3)) {
423 dollar = memchr(src + 3, '$', len - 3);
424 if (!dollar)
425 break;
426 memcpy(dst, "Id$", 3);
427 dst += 3;
428 len -= dollar + 1 - src;
429 src = dollar + 1;
430 }
431 }
432 memcpy(dst, src, len);
433 strbuf_setlen(buf, dst + len - buf->buf);
434 return 1;
435 }
436
437 static int ident_to_worktree(const char *path, const char *src, size_t len,
438 struct strbuf *buf, int ident)
439 {
440 unsigned char sha1[20];
441 char *to_free = NULL, *dollar;
442 int cnt;
443
444 if (!ident)
445 return 0;
446
447 cnt = count_ident(src, len);
448 if (!cnt)
449 return 0;
450
451 /* are we "faking" in place editing ? */
452 if (src == buf->buf)
453 to_free = strbuf_detach(buf, NULL);
454 hash_sha1_file(src, len, "blob", sha1);
455
456 strbuf_grow(buf, len + cnt * 43);
457 for (;;) {
458 /* step 1: run to the next '$' */
459 dollar = memchr(src, '$', len);
460 if (!dollar)
461 break;
462 strbuf_add(buf, src, dollar + 1 - src);
463 len -= dollar + 1 - src;
464 src = dollar + 1;
465
466 /* step 2: does it looks like a bit like Id:xxx$ or Id$ ? */
467 if (len < 3 || memcmp("Id", src, 2))
468 continue;
469
470 /* step 3: skip over Id$ or Id:xxxxx$ */
471 if (src[2] == '$') {
472 src += 3;
473 len -= 3;
474 } else if (src[2] == ':') {
475 /*
476 * It's possible that an expanded Id has crept its way into the
477 * repository, we cope with that by stripping the expansion out
478 */
479 dollar = memchr(src + 3, '$', len - 3);
480 if (!dollar) {
481 /* incomplete keyword, no more '$', so just quit the loop */
482 break;
483 }
484
485 len -= dollar + 1 - src;
486 src = dollar + 1;
487 } else {
488 /* it wasn't a "Id$" or "Id:xxxx$" */
489 continue;
490 }
491
492 /* step 4: substitute */
493 strbuf_addstr(buf, "Id: ");
494 strbuf_add(buf, sha1_to_hex(sha1), 40);
495 strbuf_addstr(buf, " $");
496 }
497 strbuf_add(buf, src, len);
498
499 free(to_free);
500 return 1;
501 }
502
503 static int git_path_check_crlf(const char *path, struct git_attr_check *check)
504 {
505 const char *value = check->value;
506
507 if (ATTR_TRUE(value))
508 return CRLF_TEXT;
509 else if (ATTR_FALSE(value))
510 return CRLF_BINARY;
511 else if (ATTR_UNSET(value))
512 ;
513 else if (!strcmp(value, "input"))
514 return CRLF_INPUT;
515 return CRLF_GUESS;
516 }
517
518 static struct convert_driver *git_path_check_convert(const char *path,
519 struct git_attr_check *check)
520 {
521 const char *value = check->value;
522 struct convert_driver *drv;
523
524 if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
525 return NULL;
526 for (drv = user_convert; drv; drv = drv->next)
527 if (!strcmp(value, drv->name))
528 return drv;
529 return NULL;
530 }
531
532 static int git_path_check_ident(const char *path, struct git_attr_check *check)
533 {
534 const char *value = check->value;
535
536 return !!ATTR_TRUE(value);
537 }
538
539 int convert_to_git(const char *path, const char *src, size_t len, struct strbuf *dst)
540 {
541 struct git_attr_check check[3];
542 int crlf = CRLF_GUESS;
543 int ident = 0, ret = 0;
544 char *filter = NULL;
545
546 setup_convert_check(check);
547 if (!git_checkattr(path, ARRAY_SIZE(check), check)) {
548 struct convert_driver *drv;
549 crlf = git_path_check_crlf(path, check + 0);
550 ident = git_path_check_ident(path, check + 1);
551 drv = git_path_check_convert(path, check + 2);
552 if (drv && drv->clean)
553 filter = drv->clean;
554 }
555
556 ret |= apply_filter(path, src, len, dst, filter);
557 if (ret) {
558 src = dst->buf;
559 len = dst->len;
560 }
561 ret |= crlf_to_git(path, src, len, dst, crlf);
562 if (ret) {
563 src = dst->buf;
564 len = dst->len;
565 }
566 return ret | ident_to_git(path, src, len, dst, ident);
567 }
568
569 int convert_to_working_tree(const char *path, const char *src, size_t len, struct strbuf *dst)
570 {
571 struct git_attr_check check[3];
572 int crlf = CRLF_GUESS;
573 int ident = 0, ret = 0;
574 char *filter = NULL;
575
576 setup_convert_check(check);
577 if (!git_checkattr(path, ARRAY_SIZE(check), check)) {
578 struct convert_driver *drv;
579 crlf = git_path_check_crlf(path, check + 0);
580 ident = git_path_check_ident(path, check + 1);
581 drv = git_path_check_convert(path, check + 2);
582 if (drv && drv->smudge)
583 filter = drv->smudge;
584 }
585
586 ret |= ident_to_worktree(path, src, len, dst, ident);
587 if (ret) {
588 src = dst->buf;
589 len = dst->len;
590 }
591 ret |= crlf_to_worktree(path, src, len, dst, crlf);
592 if (ret) {
593 src = dst->buf;
594 len = dst->len;
595 }
596 return ret | apply_filter(path, src, len, dst, filter);
597 }