]>
Commit | Line | Data |
---|---|---|
6c510bee | 1 | #include "cache.h" |
35ebfd6a | 2 | #include "attr.h" |
3fed15f5 | 3 | #include "run-command.h" |
a2b665de | 4 | #include "quote.h" |
6424c2ad | 5 | #include "sigchain.h" |
35ebfd6a | 6 | |
6c510bee LT |
7 | /* |
8 | * convert.c - convert a file when checking it out and checking it in. | |
9 | * | |
10 | * This should use the pathname to decide on whether it wants to do some | |
11 | * more interesting conversions (automatic gzip/unzip, general format | |
12 | * conversions etc etc), but by default it just does automatic CRLF<->LF | |
942e7747 | 13 | * translation when the "text" attribute or "auto_crlf" option is set. |
6c510bee LT |
14 | */ |
15 | ||
a7630bd4 TB |
16 | /* Stat bits: When BIN is set, the txt bits are unset */ |
17 | #define CONVERT_STAT_BITS_TXT_LF 0x1 | |
18 | #define CONVERT_STAT_BITS_TXT_CRLF 0x2 | |
19 | #define CONVERT_STAT_BITS_BIN 0x4 | |
20 | ||
c61dcff9 | 21 | enum crlf_action { |
df747b81 TB |
22 | CRLF_UNDEFINED, |
23 | CRLF_BINARY, | |
fd6cce9e | 24 | CRLF_TEXT, |
df747b81 TB |
25 | CRLF_TEXT_INPUT, |
26 | CRLF_TEXT_CRLF, | |
27 | CRLF_AUTO, | |
28 | CRLF_AUTO_INPUT, | |
29 | CRLF_AUTO_CRLF | |
fd6cce9e | 30 | }; |
163b9591 | 31 | |
6c510bee | 32 | struct text_stat { |
28624193 | 33 | /* NUL, CR, LF and CRLF counts */ |
6e336a53 | 34 | unsigned nul, lonecr, lonelf, crlf; |
6c510bee LT |
35 | |
36 | /* These are just approximations! */ | |
37 | unsigned printable, nonprintable; | |
38 | }; | |
39 | ||
40 | static void gather_stats(const char *buf, unsigned long size, struct text_stat *stats) | |
41 | { | |
42 | unsigned long i; | |
43 | ||
44 | memset(stats, 0, sizeof(*stats)); | |
45 | ||
46 | for (i = 0; i < size; i++) { | |
47 | unsigned char c = buf[i]; | |
48 | if (c == '\r') { | |
6e336a53 | 49 | if (i+1 < size && buf[i+1] == '\n') { |
6c510bee | 50 | stats->crlf++; |
6e336a53 TB |
51 | i++; |
52 | } else | |
53 | stats->lonecr++; | |
6c510bee LT |
54 | continue; |
55 | } | |
56 | if (c == '\n') { | |
6e336a53 | 57 | stats->lonelf++; |
6c510bee LT |
58 | continue; |
59 | } | |
60 | if (c == 127) | |
61 | /* DEL */ | |
62 | stats->nonprintable++; | |
63 | else if (c < 32) { | |
64 | switch (c) { | |
65 | /* BS, HT, ESC and FF */ | |
66 | case '\b': case '\t': case '\033': case '\014': | |
67 | stats->printable++; | |
68 | break; | |
28624193 DP |
69 | case 0: |
70 | stats->nul++; | |
71 | /* fall through */ | |
6c510bee LT |
72 | default: |
73 | stats->nonprintable++; | |
74 | } | |
75 | } | |
76 | else | |
77 | stats->printable++; | |
78 | } | |
f9dd4bf4 DK |
79 | |
80 | /* If file ends with EOF then don't count this EOF as non-printable. */ | |
81 | if (size >= 1 && buf[size-1] == '\032') | |
82 | stats->nonprintable--; | |
6c510bee LT |
83 | } |
84 | ||
85 | /* | |
86 | * The same heuristics as diff.c::mmfile_is_binary() | |
a7630bd4 | 87 | * We treat files with bare CR as binary |
6c510bee | 88 | */ |
a7630bd4 | 89 | static int convert_is_binary(unsigned long size, const struct text_stat *stats) |
6c510bee | 90 | { |
6e336a53 | 91 | if (stats->lonecr) |
a7630bd4 | 92 | return 1; |
28624193 DP |
93 | if (stats->nul) |
94 | return 1; | |
6c510bee LT |
95 | if ((stats->printable >> 7) < stats->nonprintable) |
96 | return 1; | |
6c510bee LT |
97 | return 0; |
98 | } | |
99 | ||
a7630bd4 TB |
100 | static unsigned int gather_convert_stats(const char *data, unsigned long size) |
101 | { | |
102 | struct text_stat stats; | |
6e336a53 | 103 | int ret = 0; |
a7630bd4 TB |
104 | if (!data || !size) |
105 | return 0; | |
106 | gather_stats(data, size, &stats); | |
107 | if (convert_is_binary(size, &stats)) | |
6e336a53 TB |
108 | ret |= CONVERT_STAT_BITS_BIN; |
109 | if (stats.crlf) | |
110 | ret |= CONVERT_STAT_BITS_TXT_CRLF; | |
111 | if (stats.lonelf) | |
112 | ret |= CONVERT_STAT_BITS_TXT_LF; | |
113 | ||
114 | return ret; | |
a7630bd4 TB |
115 | } |
116 | ||
117 | static const char *gather_convert_stats_ascii(const char *data, unsigned long size) | |
118 | { | |
119 | unsigned int convert_stats = gather_convert_stats(data, size); | |
120 | ||
121 | if (convert_stats & CONVERT_STAT_BITS_BIN) | |
122 | return "-text"; | |
123 | switch (convert_stats) { | |
124 | case CONVERT_STAT_BITS_TXT_LF: | |
125 | return "lf"; | |
126 | case CONVERT_STAT_BITS_TXT_CRLF: | |
127 | return "crlf"; | |
128 | case CONVERT_STAT_BITS_TXT_LF | CONVERT_STAT_BITS_TXT_CRLF: | |
129 | return "mixed"; | |
130 | default: | |
131 | return "none"; | |
132 | } | |
133 | } | |
134 | ||
135 | const char *get_cached_convert_stats_ascii(const char *path) | |
136 | { | |
137 | const char *ret; | |
138 | unsigned long sz; | |
139 | void *data = read_blob_data_from_cache(path, &sz); | |
140 | ret = gather_convert_stats_ascii(data, sz); | |
141 | free(data); | |
142 | return ret; | |
143 | } | |
144 | ||
145 | const char *get_wt_convert_stats_ascii(const char *path) | |
146 | { | |
147 | const char *ret = ""; | |
148 | struct strbuf sb = STRBUF_INIT; | |
149 | if (strbuf_read_file(&sb, path, 0) >= 0) | |
150 | ret = gather_convert_stats_ascii(sb.buf, sb.len); | |
151 | strbuf_release(&sb); | |
152 | return ret; | |
153 | } | |
154 | ||
4b4024f5 TB |
155 | static int text_eol_is_crlf(void) |
156 | { | |
157 | if (auto_crlf == AUTO_CRLF_TRUE) | |
158 | return 1; | |
159 | else if (auto_crlf == AUTO_CRLF_INPUT) | |
160 | return 0; | |
161 | if (core_eol == EOL_CRLF) | |
162 | return 1; | |
163 | if (core_eol == EOL_UNSET && EOL_NATIVE == EOL_CRLF) | |
164 | return 1; | |
165 | return 0; | |
166 | } | |
167 | ||
c61dcff9 | 168 | static enum eol output_eol(enum crlf_action crlf_action) |
f217f0e8 | 169 | { |
c61dcff9 | 170 | switch (crlf_action) { |
942e7747 EB |
171 | case CRLF_BINARY: |
172 | return EOL_UNSET; | |
df747b81 | 173 | case CRLF_TEXT_CRLF: |
942e7747 | 174 | return EOL_CRLF; |
df747b81 | 175 | case CRLF_TEXT_INPUT: |
942e7747 | 176 | return EOL_LF; |
df747b81 TB |
177 | case CRLF_UNDEFINED: |
178 | case CRLF_AUTO_CRLF: | |
179 | case CRLF_AUTO_INPUT: | |
942e7747 EB |
180 | case CRLF_TEXT: |
181 | case CRLF_AUTO: | |
df747b81 | 182 | /* fall through */ |
4b4024f5 | 183 | return text_eol_is_crlf() ? EOL_CRLF : EOL_LF; |
942e7747 | 184 | } |
df747b81 | 185 | warning("Illegal crlf_action %d\n", (int)crlf_action); |
ec70f52f | 186 | return core_eol; |
942e7747 EB |
187 | } |
188 | ||
c61dcff9 | 189 | static void check_safe_crlf(const char *path, enum crlf_action crlf_action, |
21e5ad50 SP |
190 | struct text_stat *stats, enum safe_crlf checksafe) |
191 | { | |
192 | if (!checksafe) | |
193 | return; | |
194 | ||
c61dcff9 | 195 | if (output_eol(crlf_action) == EOL_LF) { |
21e5ad50 SP |
196 | /* |
197 | * CRLFs would not be restored by checkout: | |
198 | * check if we'd remove CRLFs | |
199 | */ | |
200 | if (stats->crlf) { | |
201 | if (checksafe == SAFE_CRLF_WARN) | |
942e7747 | 202 | warning("CRLF will be replaced by LF in %s.\nThe file will have its original line endings in your working directory.", path); |
21e5ad50 SP |
203 | else /* i.e. SAFE_CRLF_FAIL */ |
204 | die("CRLF would be replaced by LF in %s.", path); | |
205 | } | |
c61dcff9 | 206 | } else if (output_eol(crlf_action) == EOL_CRLF) { |
21e5ad50 SP |
207 | /* |
208 | * CRLFs would be added by checkout: | |
209 | * check if we have "naked" LFs | |
210 | */ | |
6e336a53 | 211 | if (stats->lonelf) { |
21e5ad50 | 212 | if (checksafe == SAFE_CRLF_WARN) |
942e7747 | 213 | warning("LF will be replaced by CRLF in %s.\nThe file will have its original line endings in your working directory.", path); |
21e5ad50 SP |
214 | else /* i.e. SAFE_CRLF_FAIL */ |
215 | die("LF would be replaced by CRLF in %s", path); | |
216 | } | |
217 | } | |
218 | } | |
219 | ||
c4805393 FAG |
220 | static int has_cr_in_index(const char *path) |
221 | { | |
c4805393 | 222 | unsigned long sz; |
c4805393 FAG |
223 | void *data; |
224 | int has_cr; | |
c4805393 | 225 | |
4982fd78 LF |
226 | data = read_blob_data_from_cache(path, &sz); |
227 | if (!data) | |
c4805393 | 228 | return 0; |
c4805393 FAG |
229 | has_cr = memchr(data, '\r', sz) != NULL; |
230 | free(data); | |
231 | return has_cr; | |
232 | } | |
233 | ||
5ecd293d | 234 | static int crlf_to_git(const char *path, const char *src, size_t len, |
c61dcff9 JH |
235 | struct strbuf *buf, |
236 | enum crlf_action crlf_action, enum safe_crlf checksafe) | |
6c510bee | 237 | { |
6c510bee | 238 | struct text_stat stats; |
5ecd293d | 239 | char *dst; |
6c510bee | 240 | |
c61dcff9 | 241 | if (crlf_action == CRLF_BINARY || |
4c3b57b9 | 242 | (src && !len)) |
5ecd293d | 243 | return 0; |
6c510bee | 244 | |
4c3b57b9 JK |
245 | /* |
246 | * If we are doing a dry-run and have no source buffer, there is | |
247 | * nothing to analyze; we must assume we would convert. | |
248 | */ | |
249 | if (!buf && !src) | |
250 | return 1; | |
251 | ||
5ecd293d | 252 | gather_stats(src, len, &stats); |
6c510bee | 253 | |
df747b81 | 254 | if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) { |
a7630bd4 | 255 | if (convert_is_binary(len, &stats)) |
5ecd293d | 256 | return 0; |
c4805393 | 257 | |
df747b81 | 258 | if (crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) { |
fd6cce9e EB |
259 | /* |
260 | * If the file in the index has any CR in it, do not convert. | |
261 | * This is the new safer autocrlf handling. | |
262 | */ | |
263 | if (has_cr_in_index(path)) | |
264 | return 0; | |
265 | } | |
201ac8ef | 266 | } |
6c510bee | 267 | |
c61dcff9 | 268 | check_safe_crlf(path, crlf_action, &stats, checksafe); |
21e5ad50 | 269 | |
6e336a53 TB |
270 | /* Optimization: No CRLF? Nothing to convert, regardless. */ |
271 | if (!stats.crlf) | |
21e5ad50 SP |
272 | return 0; |
273 | ||
92ac3197 JK |
274 | /* |
275 | * At this point all of our source analysis is done, and we are sure we | |
276 | * would convert. If we are in dry-run mode, we can give an answer. | |
277 | */ | |
278 | if (!buf) | |
279 | return 1; | |
280 | ||
90d16ec0 PH |
281 | /* only grow if not in place */ |
282 | if (strbuf_avail(buf) + buf->len < len) | |
283 | strbuf_grow(buf, len - buf->len); | |
5ecd293d | 284 | dst = buf->buf; |
df747b81 | 285 | if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) { |
163b9591 JH |
286 | /* |
287 | * If we guessed, we already know we rejected a file with | |
288 | * lone CR, and we can strip a CR without looking at what | |
289 | * follow it. | |
290 | */ | |
201ac8ef | 291 | do { |
ac78e548 | 292 | unsigned char c = *src++; |
201ac8ef | 293 | if (c != '\r') |
ac78e548 | 294 | *dst++ = c; |
5ecd293d | 295 | } while (--len); |
201ac8ef JH |
296 | } else { |
297 | do { | |
ac78e548 | 298 | unsigned char c = *src++; |
5ecd293d | 299 | if (! (c == '\r' && (1 < len && *src == '\n'))) |
ac78e548 | 300 | *dst++ = c; |
5ecd293d | 301 | } while (--len); |
201ac8ef | 302 | } |
5ecd293d PH |
303 | strbuf_setlen(buf, dst - buf->buf); |
304 | return 1; | |
6c510bee LT |
305 | } |
306 | ||
5ecd293d | 307 | static int crlf_to_worktree(const char *path, const char *src, size_t len, |
c61dcff9 | 308 | struct strbuf *buf, enum crlf_action crlf_action) |
6c510bee | 309 | { |
5ecd293d | 310 | char *to_free = NULL; |
6c510bee | 311 | struct text_stat stats; |
6c510bee | 312 | |
c61dcff9 | 313 | if (!len || output_eol(crlf_action) != EOL_CRLF) |
5ecd293d | 314 | return 0; |
6c510bee | 315 | |
5ecd293d | 316 | gather_stats(src, len, &stats); |
6c510bee | 317 | |
6e336a53 TB |
318 | /* No "naked" LF? Nothing to convert, regardless. */ |
319 | if (!stats.lonelf) | |
5ecd293d | 320 | return 0; |
6c510bee | 321 | |
df747b81 TB |
322 | if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) { |
323 | if (crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) { | |
fd6cce9e EB |
324 | /* If we have any CR or CRLF line endings, we do not touch it */ |
325 | /* This is the new safer autocrlf-handling */ | |
6e336a53 | 326 | if (stats.lonecr || stats.crlf ) |
fd6cce9e EB |
327 | return 0; |
328 | } | |
c4805393 | 329 | |
a7630bd4 | 330 | if (convert_is_binary(len, &stats)) |
5ecd293d | 331 | return 0; |
201ac8ef | 332 | } |
6c510bee | 333 | |
5ecd293d PH |
334 | /* are we "faking" in place editing ? */ |
335 | if (src == buf->buf) | |
b315c5c0 | 336 | to_free = strbuf_detach(buf, NULL); |
5ecd293d | 337 | |
6e336a53 | 338 | strbuf_grow(buf, len + stats.lonelf); |
5ecd293d PH |
339 | for (;;) { |
340 | const char *nl = memchr(src, '\n', len); | |
341 | if (!nl) | |
342 | break; | |
343 | if (nl > src && nl[-1] == '\r') { | |
344 | strbuf_add(buf, src, nl + 1 - src); | |
345 | } else { | |
346 | strbuf_add(buf, src, nl - src); | |
347 | strbuf_addstr(buf, "\r\n"); | |
348 | } | |
349 | len -= nl + 1 - src; | |
350 | src = nl + 1; | |
351 | } | |
352 | strbuf_add(buf, src, len); | |
353 | ||
354 | free(to_free); | |
355 | return 1; | |
6c510bee | 356 | } |
35ebfd6a | 357 | |
546bb582 JS |
358 | struct filter_params { |
359 | const char *src; | |
360 | unsigned long size; | |
9035d75a | 361 | int fd; |
546bb582 | 362 | const char *cmd; |
a2b665de | 363 | const char *path; |
546bb582 JS |
364 | }; |
365 | ||
9035d75a | 366 | static int filter_buffer_or_fd(int in, int out, void *data) |
aa4ed402 JH |
367 | { |
368 | /* | |
369 | * Spawn cmd and feed the buffer contents through its stdin. | |
370 | */ | |
d3180279 | 371 | struct child_process child_process = CHILD_PROCESS_INIT; |
546bb582 | 372 | struct filter_params *params = (struct filter_params *)data; |
aa4ed402 | 373 | int write_err, status; |
66dbfd55 GV |
374 | const char *argv[] = { NULL, NULL }; |
375 | ||
a2b665de PW |
376 | /* apply % substitution to cmd */ |
377 | struct strbuf cmd = STRBUF_INIT; | |
378 | struct strbuf path = STRBUF_INIT; | |
379 | struct strbuf_expand_dict_entry dict[] = { | |
380 | { "f", NULL, }, | |
381 | { NULL, NULL, }, | |
382 | }; | |
383 | ||
384 | /* quote the path to preserve spaces, etc. */ | |
385 | sq_quote_buf(&path, params->path); | |
386 | dict[0].value = path.buf; | |
387 | ||
388 | /* expand all %f with the quoted path */ | |
389 | strbuf_expand(&cmd, params->cmd, strbuf_expand_dict_cb, &dict); | |
390 | strbuf_release(&path); | |
391 | ||
392 | argv[0] = cmd.buf; | |
aa4ed402 | 393 | |
dc1bfdcd | 394 | child_process.argv = argv; |
ac0ba18d | 395 | child_process.use_shell = 1; |
dc1bfdcd | 396 | child_process.in = -1; |
ae6a5609 | 397 | child_process.out = out; |
aa4ed402 | 398 | |
dc1bfdcd | 399 | if (start_command(&child_process)) |
546bb582 | 400 | return error("cannot fork to run external filter %s", params->cmd); |
aa4ed402 | 401 | |
6424c2ad JB |
402 | sigchain_push(SIGPIPE, SIG_IGN); |
403 | ||
9035d75a | 404 | if (params->src) { |
0c4dd67a JH |
405 | write_err = (write_in_full(child_process.in, |
406 | params->src, params->size) < 0); | |
407 | if (errno == EPIPE) | |
408 | write_err = 0; | |
9035d75a SP |
409 | } else { |
410 | write_err = copy_fd(params->fd, child_process.in); | |
0c4dd67a JH |
411 | if (write_err == COPY_WRITE_ERROR && errno == EPIPE) |
412 | write_err = 0; | |
9035d75a SP |
413 | } |
414 | ||
dc1bfdcd | 415 | if (close(child_process.in)) |
aa4ed402 JH |
416 | write_err = 1; |
417 | if (write_err) | |
546bb582 | 418 | error("cannot feed the input to external filter %s", params->cmd); |
aa4ed402 | 419 | |
6424c2ad JB |
420 | sigchain_pop(SIGPIPE); |
421 | ||
aa4ed402 JH |
422 | status = finish_command(&child_process); |
423 | if (status) | |
5709e036 | 424 | error("external filter %s failed %d", params->cmd, status); |
a2b665de PW |
425 | |
426 | strbuf_release(&cmd); | |
aa4ed402 JH |
427 | return (write_err || status); |
428 | } | |
429 | ||
9035d75a | 430 | static int apply_filter(const char *path, const char *src, size_t len, int fd, |
5ecd293d | 431 | struct strbuf *dst, const char *cmd) |
aa4ed402 JH |
432 | { |
433 | /* | |
434 | * Create a pipeline to have the command filter the buffer's | |
435 | * contents. | |
436 | * | |
437 | * (child --> cmd) --> us | |
438 | */ | |
546bb582 | 439 | int ret = 1; |
f285a2d7 | 440 | struct strbuf nbuf = STRBUF_INIT; |
546bb582 JS |
441 | struct async async; |
442 | struct filter_params params; | |
aa4ed402 | 443 | |
1a8630dc | 444 | if (!cmd || !*cmd) |
5ecd293d | 445 | return 0; |
aa4ed402 | 446 | |
92ac3197 JK |
447 | if (!dst) |
448 | return 1; | |
449 | ||
546bb582 | 450 | memset(&async, 0, sizeof(async)); |
9035d75a | 451 | async.proc = filter_buffer_or_fd; |
546bb582 | 452 | async.data = ¶ms; |
ae6a5609 | 453 | async.out = -1; |
546bb582 JS |
454 | params.src = src; |
455 | params.size = len; | |
9035d75a | 456 | params.fd = fd; |
546bb582 | 457 | params.cmd = cmd; |
a2b665de | 458 | params.path = path; |
aa4ed402 JH |
459 | |
460 | fflush(NULL); | |
546bb582 JS |
461 | if (start_async(&async)) |
462 | return 0; /* error was already reported */ | |
aa4ed402 | 463 | |
546bb582 | 464 | if (strbuf_read(&nbuf, async.out, len) < 0) { |
5ecd293d PH |
465 | error("read from external filter %s failed", cmd); |
466 | ret = 0; | |
aa4ed402 | 467 | } |
546bb582 | 468 | if (close(async.out)) { |
90d16ec0 | 469 | error("read from external filter %s failed", cmd); |
5ecd293d | 470 | ret = 0; |
aa4ed402 | 471 | } |
546bb582 JS |
472 | if (finish_async(&async)) { |
473 | error("external filter %s failed", cmd); | |
5ecd293d | 474 | ret = 0; |
aa4ed402 JH |
475 | } |
476 | ||
5ecd293d | 477 | if (ret) { |
90d16ec0 | 478 | strbuf_swap(dst, &nbuf); |
5ecd293d | 479 | } |
90d16ec0 | 480 | strbuf_release(&nbuf); |
5ecd293d | 481 | return ret; |
aa4ed402 JH |
482 | } |
483 | ||
484 | static struct convert_driver { | |
485 | const char *name; | |
486 | struct convert_driver *next; | |
cd8be6c9 BH |
487 | const char *smudge; |
488 | const char *clean; | |
36daaaca | 489 | int required; |
aa4ed402 JH |
490 | } *user_convert, **user_convert_tail; |
491 | ||
ef90d6d4 | 492 | static int read_convert_config(const char *var, const char *value, void *cb) |
aa4ed402 | 493 | { |
d731f0ad | 494 | const char *key, *name; |
aa4ed402 JH |
495 | int namelen; |
496 | struct convert_driver *drv; | |
497 | ||
498 | /* | |
499 | * External conversion drivers are configured using | |
500 | * "filter.<name>.variable". | |
501 | */ | |
d731f0ad | 502 | if (parse_config_key(var, "filter", &name, &namelen, &key) < 0 || !name) |
aa4ed402 | 503 | return 0; |
aa4ed402 JH |
504 | for (drv = user_convert; drv; drv = drv->next) |
505 | if (!strncmp(drv->name, name, namelen) && !drv->name[namelen]) | |
506 | break; | |
507 | if (!drv) { | |
aa4ed402 | 508 | drv = xcalloc(1, sizeof(struct convert_driver)); |
182af834 | 509 | drv->name = xmemdupz(name, namelen); |
aa4ed402 JH |
510 | *user_convert_tail = drv; |
511 | user_convert_tail = &(drv->next); | |
512 | } | |
513 | ||
aa4ed402 JH |
514 | /* |
515 | * filter.<name>.smudge and filter.<name>.clean specifies | |
516 | * the command line: | |
517 | * | |
518 | * command-line | |
519 | * | |
520 | * The command-line will not be interpolated in any way. | |
521 | */ | |
522 | ||
d731f0ad | 523 | if (!strcmp("smudge", key)) |
cd8be6c9 BH |
524 | return git_config_string(&drv->smudge, var, value); |
525 | ||
d731f0ad | 526 | if (!strcmp("clean", key)) |
cd8be6c9 | 527 | return git_config_string(&drv->clean, var, value); |
aa4ed402 | 528 | |
d731f0ad | 529 | if (!strcmp("required", key)) { |
36daaaca JB |
530 | drv->required = git_config_bool(var, value); |
531 | return 0; | |
532 | } | |
533 | ||
aa4ed402 JH |
534 | return 0; |
535 | } | |
536 | ||
3fed15f5 JH |
537 | static int count_ident(const char *cp, unsigned long size) |
538 | { | |
539 | /* | |
af9b54bb | 540 | * "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$" |
3fed15f5 JH |
541 | */ |
542 | int cnt = 0; | |
543 | char ch; | |
544 | ||
545 | while (size) { | |
546 | ch = *cp++; | |
547 | size--; | |
548 | if (ch != '$') | |
549 | continue; | |
af9b54bb | 550 | if (size < 3) |
3fed15f5 | 551 | break; |
af9b54bb | 552 | if (memcmp("Id", cp, 2)) |
3fed15f5 | 553 | continue; |
af9b54bb AP |
554 | ch = cp[2]; |
555 | cp += 3; | |
556 | size -= 3; | |
3fed15f5 | 557 | if (ch == '$') |
af9b54bb | 558 | cnt++; /* $Id$ */ |
3fed15f5 JH |
559 | if (ch != ':') |
560 | continue; | |
561 | ||
562 | /* | |
af9b54bb | 563 | * "$Id: ... "; scan up to the closing dollar sign and discard. |
3fed15f5 JH |
564 | */ |
565 | while (size) { | |
566 | ch = *cp++; | |
567 | size--; | |
568 | if (ch == '$') { | |
569 | cnt++; | |
570 | break; | |
571 | } | |
a9f3049f HG |
572 | if (ch == '\n') |
573 | break; | |
3fed15f5 JH |
574 | } |
575 | } | |
576 | return cnt; | |
577 | } | |
578 | ||
5ecd293d PH |
579 | static int ident_to_git(const char *path, const char *src, size_t len, |
580 | struct strbuf *buf, int ident) | |
3fed15f5 | 581 | { |
5ecd293d | 582 | char *dst, *dollar; |
3fed15f5 | 583 | |
4c3b57b9 | 584 | if (!ident || (src && !count_ident(src, len))) |
5ecd293d PH |
585 | return 0; |
586 | ||
92ac3197 JK |
587 | if (!buf) |
588 | return 1; | |
589 | ||
90d16ec0 PH |
590 | /* only grow if not in place */ |
591 | if (strbuf_avail(buf) + buf->len < len) | |
592 | strbuf_grow(buf, len - buf->len); | |
5ecd293d PH |
593 | dst = buf->buf; |
594 | for (;;) { | |
595 | dollar = memchr(src, '$', len); | |
596 | if (!dollar) | |
597 | break; | |
77321184 | 598 | memmove(dst, src, dollar + 1 - src); |
5ecd293d PH |
599 | dst += dollar + 1 - src; |
600 | len -= dollar + 1 - src; | |
601 | src = dollar + 1; | |
602 | ||
603 | if (len > 3 && !memcmp(src, "Id:", 3)) { | |
604 | dollar = memchr(src + 3, '$', len - 3); | |
605 | if (!dollar) | |
606 | break; | |
a9f3049f HG |
607 | if (memchr(src + 3, '\n', dollar - src - 3)) { |
608 | /* Line break before the next dollar. */ | |
609 | continue; | |
610 | } | |
611 | ||
af9b54bb AP |
612 | memcpy(dst, "Id$", 3); |
613 | dst += 3; | |
5ecd293d PH |
614 | len -= dollar + 1 - src; |
615 | src = dollar + 1; | |
3fed15f5 JH |
616 | } |
617 | } | |
77321184 | 618 | memmove(dst, src, len); |
5ecd293d PH |
619 | strbuf_setlen(buf, dst + len - buf->buf); |
620 | return 1; | |
3fed15f5 JH |
621 | } |
622 | ||
5ecd293d PH |
623 | static int ident_to_worktree(const char *path, const char *src, size_t len, |
624 | struct strbuf *buf, int ident) | |
3fed15f5 | 625 | { |
3fed15f5 | 626 | unsigned char sha1[20]; |
07814d90 | 627 | char *to_free = NULL, *dollar, *spc; |
5ecd293d | 628 | int cnt; |
3fed15f5 JH |
629 | |
630 | if (!ident) | |
5ecd293d | 631 | return 0; |
3fed15f5 | 632 | |
5ecd293d | 633 | cnt = count_ident(src, len); |
3fed15f5 | 634 | if (!cnt) |
5ecd293d | 635 | return 0; |
3fed15f5 | 636 | |
5ecd293d PH |
637 | /* are we "faking" in place editing ? */ |
638 | if (src == buf->buf) | |
b315c5c0 | 639 | to_free = strbuf_detach(buf, NULL); |
5ecd293d | 640 | hash_sha1_file(src, len, "blob", sha1); |
3fed15f5 | 641 | |
5ecd293d PH |
642 | strbuf_grow(buf, len + cnt * 43); |
643 | for (;;) { | |
644 | /* step 1: run to the next '$' */ | |
645 | dollar = memchr(src, '$', len); | |
646 | if (!dollar) | |
647 | break; | |
648 | strbuf_add(buf, src, dollar + 1 - src); | |
649 | len -= dollar + 1 - src; | |
650 | src = dollar + 1; | |
c23290d5 | 651 | |
5ecd293d PH |
652 | /* step 2: does it looks like a bit like Id:xxx$ or Id$ ? */ |
653 | if (len < 3 || memcmp("Id", src, 2)) | |
3fed15f5 JH |
654 | continue; |
655 | ||
5ecd293d PH |
656 | /* step 3: skip over Id$ or Id:xxxxx$ */ |
657 | if (src[2] == '$') { | |
658 | src += 3; | |
659 | len -= 3; | |
660 | } else if (src[2] == ':') { | |
661 | /* | |
662 | * It's possible that an expanded Id has crept its way into the | |
07814d90 HG |
663 | * repository, we cope with that by stripping the expansion out. |
664 | * This is probably not a good idea, since it will cause changes | |
665 | * on checkout, which won't go away by stash, but let's keep it | |
666 | * for git-style ids. | |
5ecd293d PH |
667 | */ |
668 | dollar = memchr(src + 3, '$', len - 3); | |
669 | if (!dollar) { | |
670 | /* incomplete keyword, no more '$', so just quit the loop */ | |
671 | break; | |
672 | } | |
c23290d5 | 673 | |
a9f3049f HG |
674 | if (memchr(src + 3, '\n', dollar - src - 3)) { |
675 | /* Line break before the next dollar. */ | |
676 | continue; | |
677 | } | |
678 | ||
07814d90 HG |
679 | spc = memchr(src + 4, ' ', dollar - src - 4); |
680 | if (spc && spc < dollar-1) { | |
681 | /* There are spaces in unexpected places. | |
682 | * This is probably an id from some other | |
683 | * versioning system. Keep it for now. | |
684 | */ | |
685 | continue; | |
686 | } | |
687 | ||
5ecd293d PH |
688 | len -= dollar + 1 - src; |
689 | src = dollar + 1; | |
690 | } else { | |
691 | /* it wasn't a "Id$" or "Id:xxxx$" */ | |
692 | continue; | |
693 | } | |
c23290d5 | 694 | |
5ecd293d PH |
695 | /* step 4: substitute */ |
696 | strbuf_addstr(buf, "Id: "); | |
697 | strbuf_add(buf, sha1_to_hex(sha1), 40); | |
698 | strbuf_addstr(buf, " $"); | |
3fed15f5 | 699 | } |
5ecd293d | 700 | strbuf_add(buf, src, len); |
3fed15f5 | 701 | |
5ecd293d PH |
702 | free(to_free); |
703 | return 1; | |
35ebfd6a JH |
704 | } |
705 | ||
92cce135 | 706 | static enum crlf_action git_path_check_crlf(struct git_attr_check *check) |
35ebfd6a | 707 | { |
6073ee85 JH |
708 | const char *value = check->value; |
709 | ||
710 | if (ATTR_TRUE(value)) | |
711 | return CRLF_TEXT; | |
712 | else if (ATTR_FALSE(value)) | |
713 | return CRLF_BINARY; | |
714 | else if (ATTR_UNSET(value)) | |
715 | ; | |
716 | else if (!strcmp(value, "input")) | |
df747b81 | 717 | return CRLF_TEXT_INPUT; |
fd6cce9e EB |
718 | else if (!strcmp(value, "auto")) |
719 | return CRLF_AUTO; | |
df747b81 | 720 | return CRLF_UNDEFINED; |
35ebfd6a JH |
721 | } |
722 | ||
92cce135 | 723 | static enum eol git_path_check_eol(struct git_attr_check *check) |
fd6cce9e EB |
724 | { |
725 | const char *value = check->value; | |
726 | ||
727 | if (ATTR_UNSET(value)) | |
728 | ; | |
729 | else if (!strcmp(value, "lf")) | |
730 | return EOL_LF; | |
731 | else if (!strcmp(value, "crlf")) | |
732 | return EOL_CRLF; | |
733 | return EOL_UNSET; | |
734 | } | |
735 | ||
92cce135 | 736 | static struct convert_driver *git_path_check_convert(struct git_attr_check *check) |
aa4ed402 JH |
737 | { |
738 | const char *value = check->value; | |
739 | struct convert_driver *drv; | |
740 | ||
741 | if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value)) | |
742 | return NULL; | |
743 | for (drv = user_convert; drv; drv = drv->next) | |
744 | if (!strcmp(value, drv->name)) | |
745 | return drv; | |
746 | return NULL; | |
747 | } | |
748 | ||
92cce135 | 749 | static int git_path_check_ident(struct git_attr_check *check) |
3fed15f5 JH |
750 | { |
751 | const char *value = check->value; | |
752 | ||
753 | return !!ATTR_TRUE(value); | |
754 | } | |
755 | ||
3bfba20d JH |
756 | struct conv_attrs { |
757 | struct convert_driver *drv; | |
bb211b4d TB |
758 | enum crlf_action attr_action; /* What attr says */ |
759 | enum crlf_action crlf_action; /* When no attr is set, use core.autocrlf */ | |
3bfba20d JH |
760 | int ident; |
761 | }; | |
762 | ||
83295964 JH |
763 | static const char *conv_attr_name[] = { |
764 | "crlf", "ident", "filter", "eol", "text", | |
765 | }; | |
766 | #define NUM_CONV_ATTRS ARRAY_SIZE(conv_attr_name) | |
767 | ||
3bfba20d | 768 | static void convert_attrs(struct conv_attrs *ca, const char *path) |
83295964 JH |
769 | { |
770 | int i; | |
771 | static struct git_attr_check ccheck[NUM_CONV_ATTRS]; | |
772 | ||
773 | if (!ccheck[0].attr) { | |
774 | for (i = 0; i < NUM_CONV_ATTRS; i++) | |
775 | ccheck[i].attr = git_attr(conv_attr_name[i]); | |
776 | user_convert_tail = &user_convert; | |
777 | git_config(read_convert_config, NULL); | |
778 | } | |
3bfba20d | 779 | |
d932f4eb | 780 | if (!git_check_attr(path, NUM_CONV_ATTRS, ccheck)) { |
bb211b4d | 781 | ca->crlf_action = git_path_check_crlf(ccheck + 4); |
df747b81 | 782 | if (ca->crlf_action == CRLF_UNDEFINED) |
92cce135 | 783 | ca->crlf_action = git_path_check_crlf(ccheck + 0); |
bb211b4d | 784 | ca->attr_action = ca->crlf_action; |
92cce135 TB |
785 | ca->ident = git_path_check_ident(ccheck + 1); |
786 | ca->drv = git_path_check_convert(ccheck + 2); | |
817a0c79 TB |
787 | if (ca->crlf_action != CRLF_BINARY) { |
788 | enum eol eol_attr = git_path_check_eol(ccheck + 3); | |
789 | if (eol_attr == EOL_LF) | |
790 | ca->crlf_action = CRLF_TEXT_INPUT; | |
791 | else if (eol_attr == EOL_CRLF) | |
792 | ca->crlf_action = CRLF_TEXT_CRLF; | |
793 | } | |
794 | ca->attr_action = ca->crlf_action; | |
3bfba20d JH |
795 | } else { |
796 | ca->drv = NULL; | |
df747b81 | 797 | ca->crlf_action = CRLF_UNDEFINED; |
3bfba20d JH |
798 | ca->ident = 0; |
799 | } | |
df747b81 TB |
800 | if (ca->crlf_action == CRLF_TEXT) |
801 | ca->crlf_action = text_eol_is_crlf() ? CRLF_TEXT_CRLF : CRLF_TEXT_INPUT; | |
802 | if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_FALSE) | |
803 | ca->crlf_action = CRLF_BINARY; | |
804 | if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_TRUE) | |
805 | ca->crlf_action = CRLF_AUTO_CRLF; | |
806 | if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_INPUT) | |
807 | ca->crlf_action = CRLF_AUTO_INPUT; | |
83295964 JH |
808 | } |
809 | ||
9035d75a SP |
810 | int would_convert_to_git_filter_fd(const char *path) |
811 | { | |
812 | struct conv_attrs ca; | |
813 | ||
814 | convert_attrs(&ca, path); | |
815 | if (!ca.drv) | |
816 | return 0; | |
817 | ||
818 | /* | |
819 | * Apply a filter to an fd only if the filter is required to succeed. | |
820 | * We must die if the filter fails, because the original data before | |
821 | * filtering is not available. | |
822 | */ | |
823 | if (!ca.drv->required) | |
824 | return 0; | |
825 | ||
826 | return apply_filter(path, NULL, 0, -1, NULL, ca.drv->clean); | |
827 | } | |
828 | ||
a7630bd4 TB |
829 | const char *get_convert_attr_ascii(const char *path) |
830 | { | |
831 | struct conv_attrs ca; | |
a7630bd4 TB |
832 | |
833 | convert_attrs(&ca, path); | |
bb211b4d | 834 | switch (ca.attr_action) { |
df747b81 | 835 | case CRLF_UNDEFINED: |
a7630bd4 TB |
836 | return ""; |
837 | case CRLF_BINARY: | |
838 | return "-text"; | |
839 | case CRLF_TEXT: | |
840 | return "text"; | |
df747b81 | 841 | case CRLF_TEXT_INPUT: |
a7630bd4 | 842 | return "text eol=lf"; |
df747b81 TB |
843 | case CRLF_TEXT_CRLF: |
844 | return "text eol=crlf"; | |
a7630bd4 TB |
845 | case CRLF_AUTO: |
846 | return "text=auto"; | |
df747b81 TB |
847 | case CRLF_AUTO_CRLF: |
848 | return "text=auto eol=crlf"; /* This is not supported yet */ | |
849 | case CRLF_AUTO_INPUT: | |
850 | return "text=auto eol=lf"; /* This is not supported yet */ | |
a7630bd4 TB |
851 | } |
852 | return ""; | |
853 | } | |
854 | ||
21e5ad50 SP |
855 | int convert_to_git(const char *path, const char *src, size_t len, |
856 | struct strbuf *dst, enum safe_crlf checksafe) | |
35ebfd6a | 857 | { |
3bfba20d | 858 | int ret = 0; |
cd8be6c9 | 859 | const char *filter = NULL; |
36daaaca | 860 | int required = 0; |
3bfba20d | 861 | struct conv_attrs ca; |
6073ee85 | 862 | |
3bfba20d | 863 | convert_attrs(&ca, path); |
36daaaca | 864 | if (ca.drv) { |
3bfba20d | 865 | filter = ca.drv->clean; |
36daaaca JB |
866 | required = ca.drv->required; |
867 | } | |
3fed15f5 | 868 | |
9035d75a | 869 | ret |= apply_filter(path, src, len, -1, dst, filter); |
36daaaca JB |
870 | if (!ret && required) |
871 | die("%s: clean filter '%s' failed", path, ca.drv->name); | |
872 | ||
92ac3197 | 873 | if (ret && dst) { |
5ecd293d PH |
874 | src = dst->buf; |
875 | len = dst->len; | |
aa4ed402 | 876 | } |
3bfba20d | 877 | ret |= crlf_to_git(path, src, len, dst, ca.crlf_action, checksafe); |
92ac3197 | 878 | if (ret && dst) { |
5ecd293d PH |
879 | src = dst->buf; |
880 | len = dst->len; | |
6073ee85 | 881 | } |
3bfba20d | 882 | return ret | ident_to_git(path, src, len, dst, ca.ident); |
35ebfd6a JH |
883 | } |
884 | ||
9035d75a SP |
885 | void convert_to_git_filter_fd(const char *path, int fd, struct strbuf *dst, |
886 | enum safe_crlf checksafe) | |
887 | { | |
888 | struct conv_attrs ca; | |
889 | convert_attrs(&ca, path); | |
890 | ||
891 | assert(ca.drv); | |
892 | assert(ca.drv->clean); | |
893 | ||
894 | if (!apply_filter(path, NULL, 0, fd, dst, ca.drv->clean)) | |
895 | die("%s: clean filter '%s' failed", path, ca.drv->name); | |
896 | ||
9035d75a SP |
897 | crlf_to_git(path, dst->buf, dst->len, dst, ca.crlf_action, checksafe); |
898 | ident_to_git(path, dst->buf, dst->len, dst, ca.ident); | |
899 | } | |
900 | ||
43dd2332 EB |
901 | static int convert_to_working_tree_internal(const char *path, const char *src, |
902 | size_t len, struct strbuf *dst, | |
903 | int normalizing) | |
35ebfd6a | 904 | { |
36daaaca | 905 | int ret = 0, ret_filter = 0; |
cd8be6c9 | 906 | const char *filter = NULL; |
36daaaca | 907 | int required = 0; |
3bfba20d | 908 | struct conv_attrs ca; |
6073ee85 | 909 | |
3bfba20d | 910 | convert_attrs(&ca, path); |
36daaaca | 911 | if (ca.drv) { |
3bfba20d | 912 | filter = ca.drv->smudge; |
36daaaca JB |
913 | required = ca.drv->required; |
914 | } | |
3fed15f5 | 915 | |
3bfba20d | 916 | ret |= ident_to_worktree(path, src, len, dst, ca.ident); |
5ecd293d PH |
917 | if (ret) { |
918 | src = dst->buf; | |
919 | len = dst->len; | |
3fed15f5 | 920 | } |
43dd2332 EB |
921 | /* |
922 | * CRLF conversion can be skipped if normalizing, unless there | |
923 | * is a smudge filter. The filter might expect CRLFs. | |
924 | */ | |
925 | if (filter || !normalizing) { | |
3bfba20d | 926 | ret |= crlf_to_worktree(path, src, len, dst, ca.crlf_action); |
43dd2332 EB |
927 | if (ret) { |
928 | src = dst->buf; | |
929 | len = dst->len; | |
930 | } | |
aa4ed402 | 931 | } |
36daaaca | 932 | |
9035d75a | 933 | ret_filter = apply_filter(path, src, len, -1, dst, filter); |
36daaaca JB |
934 | if (!ret_filter && required) |
935 | die("%s: smudge filter %s failed", path, ca.drv->name); | |
936 | ||
937 | return ret | ret_filter; | |
35ebfd6a | 938 | } |
f217f0e8 | 939 | |
43dd2332 EB |
940 | int convert_to_working_tree(const char *path, const char *src, size_t len, struct strbuf *dst) |
941 | { | |
942 | return convert_to_working_tree_internal(path, src, len, dst, 0); | |
943 | } | |
944 | ||
f217f0e8 EB |
945 | int renormalize_buffer(const char *path, const char *src, size_t len, struct strbuf *dst) |
946 | { | |
43dd2332 | 947 | int ret = convert_to_working_tree_internal(path, src, len, dst, 1); |
f217f0e8 EB |
948 | if (ret) { |
949 | src = dst->buf; | |
950 | len = dst->len; | |
951 | } | |
7356b51e | 952 | return ret | convert_to_git(path, src, len, dst, SAFE_CRLF_FALSE); |
f217f0e8 | 953 | } |
dd8e9121 | 954 | |
b6691092 JH |
955 | /***************************************************************** |
956 | * | |
749f763d | 957 | * Streaming conversion support |
b6691092 JH |
958 | * |
959 | *****************************************************************/ | |
960 | ||
961 | typedef int (*filter_fn)(struct stream_filter *, | |
962 | const char *input, size_t *isize_p, | |
963 | char *output, size_t *osize_p); | |
964 | typedef void (*free_fn)(struct stream_filter *); | |
965 | ||
966 | struct stream_filter_vtbl { | |
967 | filter_fn filter; | |
968 | free_fn free; | |
969 | }; | |
970 | ||
971 | struct stream_filter { | |
972 | struct stream_filter_vtbl *vtbl; | |
973 | }; | |
974 | ||
975 | static int null_filter_fn(struct stream_filter *filter, | |
976 | const char *input, size_t *isize_p, | |
977 | char *output, size_t *osize_p) | |
978 | { | |
4ae66704 JH |
979 | size_t count; |
980 | ||
981 | if (!input) | |
982 | return 0; /* we do not keep any states */ | |
983 | count = *isize_p; | |
b6691092 JH |
984 | if (*osize_p < count) |
985 | count = *osize_p; | |
986 | if (count) { | |
987 | memmove(output, input, count); | |
988 | *isize_p -= count; | |
989 | *osize_p -= count; | |
990 | } | |
991 | return 0; | |
992 | } | |
993 | ||
994 | static void null_free_fn(struct stream_filter *filter) | |
995 | { | |
996 | ; /* nothing -- null instances are shared */ | |
997 | } | |
998 | ||
999 | static struct stream_filter_vtbl null_vtbl = { | |
1000 | null_filter_fn, | |
1001 | null_free_fn, | |
1002 | }; | |
1003 | ||
1004 | static struct stream_filter null_filter_singleton = { | |
1005 | &null_vtbl, | |
1006 | }; | |
1007 | ||
1008 | int is_null_stream_filter(struct stream_filter *filter) | |
1009 | { | |
1010 | return filter == &null_filter_singleton; | |
1011 | } | |
1012 | ||
b84c7839 JH |
1013 | |
1014 | /* | |
1015 | * LF-to-CRLF filter | |
1016 | */ | |
284e3d28 CMN |
1017 | |
1018 | struct lf_to_crlf_filter { | |
1019 | struct stream_filter filter; | |
8496f568 JH |
1020 | unsigned has_held:1; |
1021 | char held; | |
284e3d28 CMN |
1022 | }; |
1023 | ||
e322ee38 JH |
1024 | static int lf_to_crlf_filter_fn(struct stream_filter *filter, |
1025 | const char *input, size_t *isize_p, | |
1026 | char *output, size_t *osize_p) | |
1027 | { | |
284e3d28 CMN |
1028 | size_t count, o = 0; |
1029 | struct lf_to_crlf_filter *lf_to_crlf = (struct lf_to_crlf_filter *)filter; | |
1030 | ||
8496f568 JH |
1031 | /* |
1032 | * We may be holding onto the CR to see if it is followed by a | |
1033 | * LF, in which case we would need to go to the main loop. | |
1034 | * Otherwise, just emit it to the output stream. | |
1035 | */ | |
1036 | if (lf_to_crlf->has_held && (lf_to_crlf->held != '\r' || !input)) { | |
1037 | output[o++] = lf_to_crlf->held; | |
1038 | lf_to_crlf->has_held = 0; | |
284e3d28 | 1039 | } |
e322ee38 | 1040 | |
87afe9a5 JH |
1041 | /* We are told to drain */ |
1042 | if (!input) { | |
1043 | *osize_p -= o; | |
1044 | return 0; | |
1045 | } | |
e322ee38 | 1046 | |
e322ee38 | 1047 | count = *isize_p; |
8496f568 | 1048 | if (count || lf_to_crlf->has_held) { |
284e3d28 | 1049 | size_t i; |
8496f568 JH |
1050 | int was_cr = 0; |
1051 | ||
1052 | if (lf_to_crlf->has_held) { | |
1053 | was_cr = 1; | |
1054 | lf_to_crlf->has_held = 0; | |
1055 | } | |
1056 | ||
284e3d28 | 1057 | for (i = 0; o < *osize_p && i < count; i++) { |
e322ee38 | 1058 | char ch = input[i]; |
8496f568 | 1059 | |
e322ee38 | 1060 | if (ch == '\n') { |
284e3d28 | 1061 | output[o++] = '\r'; |
8496f568 JH |
1062 | } else if (was_cr) { |
1063 | /* | |
1064 | * Previous round saw CR and it is not followed | |
1065 | * by a LF; emit the CR before processing the | |
1066 | * current character. | |
1067 | */ | |
1068 | output[o++] = '\r'; | |
e322ee38 | 1069 | } |
8496f568 JH |
1070 | |
1071 | /* | |
1072 | * We may have consumed the last output slot, | |
1073 | * in which case we need to break out of this | |
1074 | * loop; hold the current character before | |
1075 | * returning. | |
1076 | */ | |
1077 | if (*osize_p <= o) { | |
1078 | lf_to_crlf->has_held = 1; | |
1079 | lf_to_crlf->held = ch; | |
1080 | continue; /* break but increment i */ | |
1081 | } | |
1082 | ||
1083 | if (ch == '\r') { | |
1084 | was_cr = 1; | |
1085 | continue; | |
1086 | } | |
1087 | ||
1088 | was_cr = 0; | |
e322ee38 JH |
1089 | output[o++] = ch; |
1090 | } | |
1091 | ||
1092 | *osize_p -= o; | |
1093 | *isize_p -= i; | |
8496f568 JH |
1094 | |
1095 | if (!lf_to_crlf->has_held && was_cr) { | |
1096 | lf_to_crlf->has_held = 1; | |
1097 | lf_to_crlf->held = '\r'; | |
1098 | } | |
e322ee38 JH |
1099 | } |
1100 | return 0; | |
1101 | } | |
1102 | ||
284e3d28 CMN |
1103 | static void lf_to_crlf_free_fn(struct stream_filter *filter) |
1104 | { | |
1105 | free(filter); | |
1106 | } | |
1107 | ||
e322ee38 JH |
1108 | static struct stream_filter_vtbl lf_to_crlf_vtbl = { |
1109 | lf_to_crlf_filter_fn, | |
284e3d28 | 1110 | lf_to_crlf_free_fn, |
e322ee38 JH |
1111 | }; |
1112 | ||
284e3d28 CMN |
1113 | static struct stream_filter *lf_to_crlf_filter(void) |
1114 | { | |
87afe9a5 | 1115 | struct lf_to_crlf_filter *lf_to_crlf = xcalloc(1, sizeof(*lf_to_crlf)); |
e322ee38 | 1116 | |
284e3d28 | 1117 | lf_to_crlf->filter.vtbl = &lf_to_crlf_vtbl; |
284e3d28 CMN |
1118 | return (struct stream_filter *)lf_to_crlf; |
1119 | } | |
b84c7839 | 1120 | |
a265a7f9 JH |
1121 | /* |
1122 | * Cascade filter | |
1123 | */ | |
1124 | #define FILTER_BUFFER 1024 | |
1125 | struct cascade_filter { | |
1126 | struct stream_filter filter; | |
1127 | struct stream_filter *one; | |
1128 | struct stream_filter *two; | |
1129 | char buf[FILTER_BUFFER]; | |
1130 | int end, ptr; | |
1131 | }; | |
1132 | ||
1133 | static int cascade_filter_fn(struct stream_filter *filter, | |
1134 | const char *input, size_t *isize_p, | |
1135 | char *output, size_t *osize_p) | |
1136 | { | |
1137 | struct cascade_filter *cas = (struct cascade_filter *) filter; | |
1138 | size_t filled = 0; | |
1139 | size_t sz = *osize_p; | |
1140 | size_t to_feed, remaining; | |
1141 | ||
1142 | /* | |
1143 | * input -- (one) --> buf -- (two) --> output | |
1144 | */ | |
1145 | while (filled < sz) { | |
1146 | remaining = sz - filled; | |
1147 | ||
1148 | /* do we already have something to feed two with? */ | |
1149 | if (cas->ptr < cas->end) { | |
1150 | to_feed = cas->end - cas->ptr; | |
1151 | if (stream_filter(cas->two, | |
1152 | cas->buf + cas->ptr, &to_feed, | |
1153 | output + filled, &remaining)) | |
1154 | return -1; | |
1155 | cas->ptr += (cas->end - cas->ptr) - to_feed; | |
1156 | filled = sz - remaining; | |
1157 | continue; | |
1158 | } | |
1159 | ||
1160 | /* feed one from upstream and have it emit into our buffer */ | |
1161 | to_feed = input ? *isize_p : 0; | |
1162 | if (input && !to_feed) | |
1163 | break; | |
1164 | remaining = sizeof(cas->buf); | |
1165 | if (stream_filter(cas->one, | |
1166 | input, &to_feed, | |
1167 | cas->buf, &remaining)) | |
1168 | return -1; | |
1169 | cas->end = sizeof(cas->buf) - remaining; | |
1170 | cas->ptr = 0; | |
1171 | if (input) { | |
1172 | size_t fed = *isize_p - to_feed; | |
1173 | *isize_p -= fed; | |
1174 | input += fed; | |
1175 | } | |
1176 | ||
1177 | /* do we know that we drained one completely? */ | |
1178 | if (input || cas->end) | |
1179 | continue; | |
1180 | ||
1181 | /* tell two to drain; we have nothing more to give it */ | |
1182 | to_feed = 0; | |
1183 | remaining = sz - filled; | |
1184 | if (stream_filter(cas->two, | |
1185 | NULL, &to_feed, | |
1186 | output + filled, &remaining)) | |
1187 | return -1; | |
1188 | if (remaining == (sz - filled)) | |
1189 | break; /* completely drained two */ | |
1190 | filled = sz - remaining; | |
1191 | } | |
1192 | *osize_p -= filled; | |
1193 | return 0; | |
1194 | } | |
1195 | ||
1196 | static void cascade_free_fn(struct stream_filter *filter) | |
1197 | { | |
1198 | struct cascade_filter *cas = (struct cascade_filter *)filter; | |
1199 | free_stream_filter(cas->one); | |
1200 | free_stream_filter(cas->two); | |
1201 | free(filter); | |
1202 | } | |
1203 | ||
1204 | static struct stream_filter_vtbl cascade_vtbl = { | |
1205 | cascade_filter_fn, | |
1206 | cascade_free_fn, | |
1207 | }; | |
1208 | ||
1209 | static struct stream_filter *cascade_filter(struct stream_filter *one, | |
1210 | struct stream_filter *two) | |
1211 | { | |
1212 | struct cascade_filter *cascade; | |
1213 | ||
1214 | if (!one || is_null_stream_filter(one)) | |
1215 | return two; | |
1216 | if (!two || is_null_stream_filter(two)) | |
1217 | return one; | |
1218 | ||
1219 | cascade = xmalloc(sizeof(*cascade)); | |
1220 | cascade->one = one; | |
1221 | cascade->two = two; | |
1222 | cascade->end = cascade->ptr = 0; | |
1223 | cascade->filter.vtbl = &cascade_vtbl; | |
1224 | return (struct stream_filter *)cascade; | |
1225 | } | |
1226 | ||
b84c7839 JH |
1227 | /* |
1228 | * ident filter | |
1229 | */ | |
1230 | #define IDENT_DRAINING (-1) | |
1231 | #define IDENT_SKIPPING (-2) | |
1232 | struct ident_filter { | |
1233 | struct stream_filter filter; | |
1234 | struct strbuf left; | |
1235 | int state; | |
1236 | char ident[45]; /* ": x40 $" */ | |
1237 | }; | |
1238 | ||
1239 | static int is_foreign_ident(const char *str) | |
1240 | { | |
1241 | int i; | |
1242 | ||
ae021d87 | 1243 | if (!skip_prefix(str, "$Id: ", &str)) |
b84c7839 | 1244 | return 0; |
ae021d87 | 1245 | for (i = 0; str[i]; i++) { |
b84c7839 JH |
1246 | if (isspace(str[i]) && str[i+1] != '$') |
1247 | return 1; | |
1248 | } | |
1249 | return 0; | |
1250 | } | |
1251 | ||
1252 | static void ident_drain(struct ident_filter *ident, char **output_p, size_t *osize_p) | |
1253 | { | |
1254 | size_t to_drain = ident->left.len; | |
1255 | ||
1256 | if (*osize_p < to_drain) | |
1257 | to_drain = *osize_p; | |
1258 | if (to_drain) { | |
1259 | memcpy(*output_p, ident->left.buf, to_drain); | |
1260 | strbuf_remove(&ident->left, 0, to_drain); | |
1261 | *output_p += to_drain; | |
1262 | *osize_p -= to_drain; | |
1263 | } | |
1264 | if (!ident->left.len) | |
1265 | ident->state = 0; | |
1266 | } | |
1267 | ||
1268 | static int ident_filter_fn(struct stream_filter *filter, | |
1269 | const char *input, size_t *isize_p, | |
1270 | char *output, size_t *osize_p) | |
1271 | { | |
1272 | struct ident_filter *ident = (struct ident_filter *)filter; | |
1273 | static const char head[] = "$Id"; | |
1274 | ||
1275 | if (!input) { | |
1276 | /* drain upon eof */ | |
1277 | switch (ident->state) { | |
1278 | default: | |
1279 | strbuf_add(&ident->left, head, ident->state); | |
1280 | case IDENT_SKIPPING: | |
1281 | /* fallthru */ | |
1282 | case IDENT_DRAINING: | |
1283 | ident_drain(ident, &output, osize_p); | |
1284 | } | |
1285 | return 0; | |
1286 | } | |
1287 | ||
1288 | while (*isize_p || (ident->state == IDENT_DRAINING)) { | |
1289 | int ch; | |
1290 | ||
1291 | if (ident->state == IDENT_DRAINING) { | |
1292 | ident_drain(ident, &output, osize_p); | |
1293 | if (!*osize_p) | |
1294 | break; | |
1295 | continue; | |
1296 | } | |
1297 | ||
1298 | ch = *(input++); | |
1299 | (*isize_p)--; | |
1300 | ||
1301 | if (ident->state == IDENT_SKIPPING) { | |
1302 | /* | |
1303 | * Skipping until '$' or LF, but keeping them | |
1304 | * in case it is a foreign ident. | |
1305 | */ | |
1306 | strbuf_addch(&ident->left, ch); | |
1307 | if (ch != '\n' && ch != '$') | |
1308 | continue; | |
1309 | if (ch == '$' && !is_foreign_ident(ident->left.buf)) { | |
1310 | strbuf_setlen(&ident->left, sizeof(head) - 1); | |
1311 | strbuf_addstr(&ident->left, ident->ident); | |
1312 | } | |
1313 | ident->state = IDENT_DRAINING; | |
1314 | continue; | |
1315 | } | |
1316 | ||
1317 | if (ident->state < sizeof(head) && | |
1318 | head[ident->state] == ch) { | |
1319 | ident->state++; | |
1320 | continue; | |
1321 | } | |
1322 | ||
1323 | if (ident->state) | |
1324 | strbuf_add(&ident->left, head, ident->state); | |
1325 | if (ident->state == sizeof(head) - 1) { | |
1326 | if (ch != ':' && ch != '$') { | |
1327 | strbuf_addch(&ident->left, ch); | |
1328 | ident->state = 0; | |
1329 | continue; | |
1330 | } | |
1331 | ||
1332 | if (ch == ':') { | |
1333 | strbuf_addch(&ident->left, ch); | |
1334 | ident->state = IDENT_SKIPPING; | |
1335 | } else { | |
1336 | strbuf_addstr(&ident->left, ident->ident); | |
1337 | ident->state = IDENT_DRAINING; | |
1338 | } | |
1339 | continue; | |
1340 | } | |
1341 | ||
1342 | strbuf_addch(&ident->left, ch); | |
1343 | ident->state = IDENT_DRAINING; | |
1344 | } | |
1345 | return 0; | |
1346 | } | |
1347 | ||
1348 | static void ident_free_fn(struct stream_filter *filter) | |
1349 | { | |
1350 | struct ident_filter *ident = (struct ident_filter *)filter; | |
1351 | strbuf_release(&ident->left); | |
1352 | free(filter); | |
1353 | } | |
1354 | ||
1355 | static struct stream_filter_vtbl ident_vtbl = { | |
1356 | ident_filter_fn, | |
1357 | ident_free_fn, | |
1358 | }; | |
1359 | ||
1360 | static struct stream_filter *ident_filter(const unsigned char *sha1) | |
1361 | { | |
1362 | struct ident_filter *ident = xmalloc(sizeof(*ident)); | |
1363 | ||
5096d490 JK |
1364 | xsnprintf(ident->ident, sizeof(ident->ident), |
1365 | ": %s $", sha1_to_hex(sha1)); | |
b84c7839 JH |
1366 | strbuf_init(&ident->left, 0); |
1367 | ident->filter.vtbl = &ident_vtbl; | |
1368 | ident->state = 0; | |
1369 | return (struct stream_filter *)ident; | |
1370 | } | |
1371 | ||
dd8e9121 | 1372 | /* |
b6691092 JH |
1373 | * Return an appropriately constructed filter for the path, or NULL if |
1374 | * the contents cannot be filtered without reading the whole thing | |
1375 | * in-core. | |
1376 | * | |
1377 | * Note that you would be crazy to set CRLF, smuge/clean or ident to a | |
1378 | * large binary blob you would want us not to slurp into the memory! | |
dd8e9121 | 1379 | */ |
b6691092 | 1380 | struct stream_filter *get_stream_filter(const char *path, const unsigned char *sha1) |
dd8e9121 JH |
1381 | { |
1382 | struct conv_attrs ca; | |
1383 | enum crlf_action crlf_action; | |
b84c7839 | 1384 | struct stream_filter *filter = NULL; |
dd8e9121 JH |
1385 | |
1386 | convert_attrs(&ca, path); | |
1387 | ||
b84c7839 JH |
1388 | if (ca.drv && (ca.drv->smudge || ca.drv->clean)) |
1389 | return filter; | |
1390 | ||
1391 | if (ca.ident) | |
1392 | filter = ident_filter(sha1); | |
dd8e9121 | 1393 | |
bb211b4d | 1394 | crlf_action = ca.crlf_action; |
b84c7839 | 1395 | |
df747b81 TB |
1396 | if ((crlf_action == CRLF_BINARY) || |
1397 | crlf_action == CRLF_AUTO_INPUT || | |
1398 | (crlf_action == CRLF_TEXT_INPUT)) | |
a265a7f9 JH |
1399 | filter = cascade_filter(filter, &null_filter_singleton); |
1400 | ||
1401 | else if (output_eol(crlf_action) == EOL_CRLF && | |
df747b81 | 1402 | !(crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_CRLF)) |
284e3d28 | 1403 | filter = cascade_filter(filter, lf_to_crlf_filter()); |
e322ee38 | 1404 | |
b84c7839 | 1405 | return filter; |
b6691092 JH |
1406 | } |
1407 | ||
1408 | void free_stream_filter(struct stream_filter *filter) | |
1409 | { | |
1410 | filter->vtbl->free(filter); | |
1411 | } | |
1412 | ||
1413 | int stream_filter(struct stream_filter *filter, | |
1414 | const char *input, size_t *isize_p, | |
1415 | char *output, size_t *osize_p) | |
1416 | { | |
1417 | return filter->vtbl->filter(filter, input, isize_p, output, osize_p); | |
dd8e9121 | 1418 | } |