| 1 | #define USE_THE_REPOSITORY_VARIABLE |
| 2 | #define DISABLE_SIGN_COMPARE_WARNINGS |
| 3 | |
| 4 | #include "git-compat-util.h" |
| 5 | #include "advice.h" |
| 6 | #include "config.h" |
| 7 | #include "convert.h" |
| 8 | #include "copy.h" |
| 9 | #include "gettext.h" |
| 10 | #include "hex.h" |
| 11 | #include "object-file.h" |
| 12 | #include "attr.h" |
| 13 | #include "run-command.h" |
| 14 | #include "quote.h" |
| 15 | #include "read-cache-ll.h" |
| 16 | #include "sigchain.h" |
| 17 | #include "pkt-line.h" |
| 18 | #include "sub-process.h" |
| 19 | #include "trace.h" |
| 20 | #include "utf8.h" |
| 21 | #include "merge-ll.h" |
| 22 | |
| 23 | /* |
| 24 | * convert.c - convert a file when checking it out and checking it in. |
| 25 | * |
| 26 | * This should use the pathname to decide on whether it wants to do some |
| 27 | * more interesting conversions (automatic gzip/unzip, general format |
| 28 | * conversions etc etc), but by default it just does automatic CRLF<->LF |
| 29 | * translation when the "text" attribute or "auto_crlf" option is set. |
| 30 | */ |
| 31 | |
| 32 | /* Stat bits: When BIN is set, the txt bits are unset */ |
| 33 | #define CONVERT_STAT_BITS_TXT_LF 0x1 |
| 34 | #define CONVERT_STAT_BITS_TXT_CRLF 0x2 |
| 35 | #define CONVERT_STAT_BITS_BIN 0x4 |
| 36 | |
| 37 | struct text_stat { |
| 38 | /* NUL, CR, LF and CRLF counts */ |
| 39 | unsigned nul, lonecr, lonelf, crlf; |
| 40 | |
| 41 | /* These are just approximations! */ |
| 42 | unsigned printable, nonprintable; |
| 43 | }; |
| 44 | |
| 45 | static void gather_stats(const char *buf, unsigned long size, struct text_stat *stats) |
| 46 | { |
| 47 | unsigned long i; |
| 48 | |
| 49 | memset(stats, 0, sizeof(*stats)); |
| 50 | |
| 51 | for (i = 0; i < size; i++) { |
| 52 | unsigned char c = buf[i]; |
| 53 | if (c == '\r') { |
| 54 | if (i+1 < size && buf[i+1] == '\n') { |
| 55 | stats->crlf++; |
| 56 | i++; |
| 57 | } else |
| 58 | stats->lonecr++; |
| 59 | continue; |
| 60 | } |
| 61 | if (c == '\n') { |
| 62 | stats->lonelf++; |
| 63 | continue; |
| 64 | } |
| 65 | if (c == 127) |
| 66 | /* DEL */ |
| 67 | stats->nonprintable++; |
| 68 | else if (c < 32) { |
| 69 | switch (c) { |
| 70 | /* BS, HT, ESC and FF */ |
| 71 | case '\b': case '\t': case '\033': case '\014': |
| 72 | stats->printable++; |
| 73 | break; |
| 74 | case 0: |
| 75 | stats->nul++; |
| 76 | /* fall through */ |
| 77 | default: |
| 78 | stats->nonprintable++; |
| 79 | } |
| 80 | } |
| 81 | else |
| 82 | stats->printable++; |
| 83 | } |
| 84 | |
| 85 | /* If file ends with EOF then don't count this EOF as non-printable. */ |
| 86 | if (size >= 1 && buf[size-1] == '\032') |
| 87 | stats->nonprintable--; |
| 88 | } |
| 89 | |
| 90 | /* |
| 91 | * The same heuristics as diff.c::mmfile_is_binary() |
| 92 | * We treat files with bare CR as binary |
| 93 | */ |
| 94 | static int convert_is_binary(const struct text_stat *stats) |
| 95 | { |
| 96 | if (stats->lonecr) |
| 97 | return 1; |
| 98 | if (stats->nul) |
| 99 | return 1; |
| 100 | if ((stats->printable >> 7) < stats->nonprintable) |
| 101 | return 1; |
| 102 | return 0; |
| 103 | } |
| 104 | |
| 105 | static unsigned int gather_convert_stats(const char *data, unsigned long size) |
| 106 | { |
| 107 | struct text_stat stats; |
| 108 | int ret = 0; |
| 109 | if (!data || !size) |
| 110 | return 0; |
| 111 | gather_stats(data, size, &stats); |
| 112 | if (convert_is_binary(&stats)) |
| 113 | ret |= CONVERT_STAT_BITS_BIN; |
| 114 | if (stats.crlf) |
| 115 | ret |= CONVERT_STAT_BITS_TXT_CRLF; |
| 116 | if (stats.lonelf) |
| 117 | ret |= CONVERT_STAT_BITS_TXT_LF; |
| 118 | |
| 119 | return ret; |
| 120 | } |
| 121 | |
| 122 | static const char *gather_convert_stats_ascii(const char *data, unsigned long size) |
| 123 | { |
| 124 | unsigned int convert_stats = gather_convert_stats(data, size); |
| 125 | |
| 126 | if (convert_stats & CONVERT_STAT_BITS_BIN) |
| 127 | return "-text"; |
| 128 | switch (convert_stats) { |
| 129 | case CONVERT_STAT_BITS_TXT_LF: |
| 130 | return "lf"; |
| 131 | case CONVERT_STAT_BITS_TXT_CRLF: |
| 132 | return "crlf"; |
| 133 | case CONVERT_STAT_BITS_TXT_LF | CONVERT_STAT_BITS_TXT_CRLF: |
| 134 | return "mixed"; |
| 135 | default: |
| 136 | return "none"; |
| 137 | } |
| 138 | } |
| 139 | |
| 140 | const char *get_cached_convert_stats_ascii(struct index_state *istate, |
| 141 | const char *path) |
| 142 | { |
| 143 | const char *ret; |
| 144 | unsigned long sz; |
| 145 | void *data = read_blob_data_from_index(istate, path, &sz); |
| 146 | ret = gather_convert_stats_ascii(data, sz); |
| 147 | free(data); |
| 148 | return ret; |
| 149 | } |
| 150 | |
| 151 | const char *get_wt_convert_stats_ascii(const char *path) |
| 152 | { |
| 153 | const char *ret = ""; |
| 154 | struct strbuf sb = STRBUF_INIT; |
| 155 | if (strbuf_read_file(&sb, path, 0) >= 0) |
| 156 | ret = gather_convert_stats_ascii(sb.buf, sb.len); |
| 157 | strbuf_release(&sb); |
| 158 | return ret; |
| 159 | } |
| 160 | |
| 161 | static int text_eol_is_crlf(void) |
| 162 | { |
| 163 | if (auto_crlf == AUTO_CRLF_TRUE) |
| 164 | return 1; |
| 165 | else if (auto_crlf == AUTO_CRLF_INPUT) |
| 166 | return 0; |
| 167 | if (core_eol == EOL_CRLF) |
| 168 | return 1; |
| 169 | if (core_eol == EOL_UNSET && EOL_NATIVE == EOL_CRLF) |
| 170 | return 1; |
| 171 | return 0; |
| 172 | } |
| 173 | |
| 174 | static enum eol output_eol(enum convert_crlf_action crlf_action) |
| 175 | { |
| 176 | switch (crlf_action) { |
| 177 | case CRLF_BINARY: |
| 178 | return EOL_UNSET; |
| 179 | case CRLF_TEXT_CRLF: |
| 180 | return EOL_CRLF; |
| 181 | case CRLF_TEXT_INPUT: |
| 182 | return EOL_LF; |
| 183 | case CRLF_UNDEFINED: |
| 184 | case CRLF_AUTO_CRLF: |
| 185 | return EOL_CRLF; |
| 186 | case CRLF_AUTO_INPUT: |
| 187 | return EOL_LF; |
| 188 | case CRLF_TEXT: |
| 189 | case CRLF_AUTO: |
| 190 | /* fall through */ |
| 191 | return text_eol_is_crlf() ? EOL_CRLF : EOL_LF; |
| 192 | } |
| 193 | warning(_("illegal crlf_action %d"), (int)crlf_action); |
| 194 | return core_eol; |
| 195 | } |
| 196 | |
| 197 | static void check_global_conv_flags_eol(const char *path, |
| 198 | struct text_stat *old_stats, struct text_stat *new_stats, |
| 199 | int conv_flags) |
| 200 | { |
| 201 | if (old_stats->crlf && !new_stats->crlf ) { |
| 202 | /* |
| 203 | * CRLFs would not be restored by checkout |
| 204 | */ |
| 205 | if (conv_flags & CONV_EOL_RNDTRP_DIE) |
| 206 | die(_("CRLF would be replaced by LF in %s"), path); |
| 207 | else if (conv_flags & CONV_EOL_RNDTRP_WARN) |
| 208 | warning(_("in the working copy of '%s', CRLF will be" |
| 209 | " replaced by LF the next time Git touches" |
| 210 | " it"), path); |
| 211 | } else if (old_stats->lonelf && !new_stats->lonelf ) { |
| 212 | /* |
| 213 | * CRLFs would be added by checkout |
| 214 | */ |
| 215 | if (conv_flags & CONV_EOL_RNDTRP_DIE) |
| 216 | die(_("LF would be replaced by CRLF in %s"), path); |
| 217 | else if (conv_flags & CONV_EOL_RNDTRP_WARN) |
| 218 | warning(_("in the working copy of '%s', LF will be" |
| 219 | " replaced by CRLF the next time Git touches" |
| 220 | " it"), path); |
| 221 | } |
| 222 | } |
| 223 | |
| 224 | static int has_crlf_in_index(struct index_state *istate, const char *path) |
| 225 | { |
| 226 | unsigned long sz; |
| 227 | void *data; |
| 228 | const char *crp; |
| 229 | int has_crlf = 0; |
| 230 | |
| 231 | data = read_blob_data_from_index(istate, path, &sz); |
| 232 | if (!data) |
| 233 | return 0; |
| 234 | |
| 235 | crp = memchr(data, '\r', sz); |
| 236 | if (crp) { |
| 237 | unsigned int ret_stats; |
| 238 | ret_stats = gather_convert_stats(data, sz); |
| 239 | if (!(ret_stats & CONVERT_STAT_BITS_BIN) && |
| 240 | (ret_stats & CONVERT_STAT_BITS_TXT_CRLF)) |
| 241 | has_crlf = 1; |
| 242 | } |
| 243 | free(data); |
| 244 | return has_crlf; |
| 245 | } |
| 246 | |
| 247 | static int will_convert_lf_to_crlf(struct text_stat *stats, |
| 248 | enum convert_crlf_action crlf_action) |
| 249 | { |
| 250 | if (output_eol(crlf_action) != EOL_CRLF) |
| 251 | return 0; |
| 252 | /* No "naked" LF? Nothing to convert, regardless. */ |
| 253 | if (!stats->lonelf) |
| 254 | return 0; |
| 255 | |
| 256 | if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) { |
| 257 | /* If we have any CR or CRLF line endings, we do not touch it */ |
| 258 | /* This is the new safer autocrlf-handling */ |
| 259 | if (stats->lonecr || stats->crlf) |
| 260 | return 0; |
| 261 | |
| 262 | if (convert_is_binary(stats)) |
| 263 | return 0; |
| 264 | } |
| 265 | return 1; |
| 266 | |
| 267 | } |
| 268 | |
| 269 | static int validate_encoding(const char *path, const char *enc, |
| 270 | const char *data, size_t len, int die_on_error) |
| 271 | { |
| 272 | const char *stripped; |
| 273 | |
| 274 | /* We only check for UTF here as UTF?? can be an alias for UTF-?? */ |
| 275 | if (skip_iprefix(enc, "UTF", &stripped)) { |
| 276 | skip_prefix(stripped, "-", &stripped); |
| 277 | |
| 278 | /* |
| 279 | * Check for detectable errors in UTF encodings |
| 280 | */ |
| 281 | if (has_prohibited_utf_bom(enc, data, len)) { |
| 282 | const char *error_msg = _( |
| 283 | "BOM is prohibited in '%s' if encoded as %s"); |
| 284 | /* |
| 285 | * This advice is shown for UTF-??BE and UTF-??LE encodings. |
| 286 | * We cut off the last two characters of the encoding name |
| 287 | * to generate the encoding name suitable for BOMs. |
| 288 | */ |
| 289 | const char *advise_msg = _( |
| 290 | "The file '%s' contains a byte order " |
| 291 | "mark (BOM). Please use UTF-%.*s as " |
| 292 | "working-tree-encoding."); |
| 293 | int stripped_len = strlen(stripped) - strlen("BE"); |
| 294 | advise(advise_msg, path, stripped_len, stripped); |
| 295 | if (die_on_error) |
| 296 | die(error_msg, path, enc); |
| 297 | else { |
| 298 | return error(error_msg, path, enc); |
| 299 | } |
| 300 | |
| 301 | } else if (is_missing_required_utf_bom(enc, data, len)) { |
| 302 | const char *error_msg = _( |
| 303 | "BOM is required in '%s' if encoded as %s"); |
| 304 | const char *advise_msg = _( |
| 305 | "The file '%s' is missing a byte order " |
| 306 | "mark (BOM). Please use UTF-%sBE or UTF-%sLE " |
| 307 | "(depending on the byte order) as " |
| 308 | "working-tree-encoding."); |
| 309 | advise(advise_msg, path, stripped, stripped); |
| 310 | if (die_on_error) |
| 311 | die(error_msg, path, enc); |
| 312 | else { |
| 313 | return error(error_msg, path, enc); |
| 314 | } |
| 315 | } |
| 316 | |
| 317 | } |
| 318 | return 0; |
| 319 | } |
| 320 | |
| 321 | static void trace_encoding(const char *context, const char *path, |
| 322 | const char *encoding, const char *buf, size_t len) |
| 323 | { |
| 324 | static struct trace_key coe = TRACE_KEY_INIT(WORKING_TREE_ENCODING); |
| 325 | struct strbuf trace = STRBUF_INIT; |
| 326 | int i; |
| 327 | |
| 328 | if (!trace_want(&coe)) |
| 329 | return; |
| 330 | |
| 331 | strbuf_addf(&trace, "%s (%s, considered %s):\n", context, path, encoding); |
| 332 | for (i = 0; i < len && buf; ++i) { |
| 333 | strbuf_addf( |
| 334 | &trace, "| \033[2m%2i:\033[0m %2x \033[2m%c\033[0m%c", |
| 335 | i, |
| 336 | (unsigned char) buf[i], |
| 337 | (buf[i] > 32 && buf[i] < 127 ? buf[i] : ' '), |
| 338 | ((i+1) % 8 && (i+1) < len ? ' ' : '\n') |
| 339 | ); |
| 340 | } |
| 341 | strbuf_addchars(&trace, '\n', 1); |
| 342 | |
| 343 | trace_strbuf(&coe, &trace); |
| 344 | strbuf_release(&trace); |
| 345 | } |
| 346 | |
| 347 | static int check_roundtrip(const char *enc_name) |
| 348 | { |
| 349 | /* |
| 350 | * check_roundtrip_encoding contains a string of comma and/or |
| 351 | * space separated encodings (eg. "UTF-16, ASCII, CP1125"). |
| 352 | * Search for the given encoding in that string. |
| 353 | */ |
| 354 | const char *encoding = check_roundtrip_encoding ? |
| 355 | check_roundtrip_encoding : "SHIFT-JIS"; |
| 356 | const char *found = strcasestr(encoding, enc_name); |
| 357 | const char *next; |
| 358 | int len; |
| 359 | if (!found) |
| 360 | return 0; |
| 361 | next = found + strlen(enc_name); |
| 362 | len = strlen(encoding); |
| 363 | return (found && ( |
| 364 | /* |
| 365 | * Check that the found encoding is at the beginning of |
| 366 | * encoding or that it is prefixed with a space or |
| 367 | * comma. |
| 368 | */ |
| 369 | found == encoding || ( |
| 370 | (isspace(found[-1]) || found[-1] == ',') |
| 371 | ) |
| 372 | ) && ( |
| 373 | /* |
| 374 | * Check that the found encoding is at the end of |
| 375 | * encoding or that it is suffixed with a space |
| 376 | * or comma. |
| 377 | */ |
| 378 | next == encoding + len || ( |
| 379 | next < encoding + len && |
| 380 | (isspace(next[0]) || next[0] == ',') |
| 381 | ) |
| 382 | )); |
| 383 | } |
| 384 | |
| 385 | static const char *default_encoding = "UTF-8"; |
| 386 | |
| 387 | static int encode_to_git(const char *path, const char *src, size_t src_len, |
| 388 | struct strbuf *buf, const char *enc, int conv_flags) |
| 389 | { |
| 390 | char *dst; |
| 391 | size_t dst_len; |
| 392 | int die_on_error = conv_flags & CONV_WRITE_OBJECT; |
| 393 | |
| 394 | /* |
| 395 | * No encoding is specified or there is nothing to encode. |
| 396 | * Tell the caller that the content was not modified. |
| 397 | */ |
| 398 | if (!enc || (src && !src_len)) |
| 399 | return 0; |
| 400 | |
| 401 | /* |
| 402 | * Looks like we got called from "would_convert_to_git()". |
| 403 | * This means Git wants to know if it would encode (= modify!) |
| 404 | * the content. Let's answer with "yes", since an encoding was |
| 405 | * specified. |
| 406 | */ |
| 407 | if (!buf && !src) |
| 408 | return 1; |
| 409 | |
| 410 | if (validate_encoding(path, enc, src, src_len, die_on_error)) |
| 411 | return 0; |
| 412 | |
| 413 | trace_encoding("source", path, enc, src, src_len); |
| 414 | dst = reencode_string_len(src, src_len, default_encoding, enc, |
| 415 | &dst_len); |
| 416 | if (!dst) { |
| 417 | /* |
| 418 | * We could add the blob "as-is" to Git. However, on checkout |
| 419 | * we would try to re-encode to the original encoding. This |
| 420 | * would fail and we would leave the user with a messed-up |
| 421 | * working tree. Let's try to avoid this by screaming loud. |
| 422 | */ |
| 423 | const char* msg = _("failed to encode '%s' from %s to %s"); |
| 424 | if (die_on_error) |
| 425 | die(msg, path, enc, default_encoding); |
| 426 | else { |
| 427 | error(msg, path, enc, default_encoding); |
| 428 | return 0; |
| 429 | } |
| 430 | } |
| 431 | trace_encoding("destination", path, default_encoding, dst, dst_len); |
| 432 | |
| 433 | /* |
| 434 | * UTF supports lossless conversion round tripping [1] and conversions |
| 435 | * between UTF and other encodings are mostly round trip safe as |
| 436 | * Unicode aims to be a superset of all other character encodings. |
| 437 | * However, certain encodings (e.g. SHIFT-JIS) are known to have round |
| 438 | * trip issues [2]. Check the round trip conversion for all encodings |
| 439 | * listed in core.checkRoundtripEncoding. |
| 440 | * |
| 441 | * The round trip check is only performed if content is written to Git. |
| 442 | * This ensures that no information is lost during conversion to/from |
| 443 | * the internal UTF-8 representation. |
| 444 | * |
| 445 | * Please note, the code below is not tested because I was not able to |
| 446 | * generate a faulty round trip without an iconv error. Iconv errors |
| 447 | * are already caught above. |
| 448 | * |
| 449 | * [1] http://unicode.org/faq/utf_bom.html#gen2 |
| 450 | * [2] https://support.microsoft.com/en-us/help/170559/prb-conversion-problem-between-shift-jis-and-unicode |
| 451 | */ |
| 452 | if (die_on_error && check_roundtrip(enc)) { |
| 453 | char *re_src; |
| 454 | size_t re_src_len; |
| 455 | |
| 456 | re_src = reencode_string_len(dst, dst_len, |
| 457 | enc, default_encoding, |
| 458 | &re_src_len); |
| 459 | |
| 460 | trace_printf("Checking roundtrip encoding for %s...\n", enc); |
| 461 | trace_encoding("reencoded source", path, enc, |
| 462 | re_src, re_src_len); |
| 463 | |
| 464 | if (!re_src || src_len != re_src_len || |
| 465 | memcmp(src, re_src, src_len)) { |
| 466 | const char* msg = _("encoding '%s' from %s to %s and " |
| 467 | "back is not the same"); |
| 468 | die(msg, path, enc, default_encoding); |
| 469 | } |
| 470 | |
| 471 | free(re_src); |
| 472 | } |
| 473 | |
| 474 | strbuf_attach(buf, dst, dst_len, dst_len + 1); |
| 475 | return 1; |
| 476 | } |
| 477 | |
| 478 | static int encode_to_worktree(const char *path, const char *src, size_t src_len, |
| 479 | struct strbuf *buf, const char *enc) |
| 480 | { |
| 481 | char *dst; |
| 482 | size_t dst_len; |
| 483 | |
| 484 | /* |
| 485 | * No encoding is specified or there is nothing to encode. |
| 486 | * Tell the caller that the content was not modified. |
| 487 | */ |
| 488 | if (!enc || (src && !src_len)) |
| 489 | return 0; |
| 490 | |
| 491 | dst = reencode_string_len(src, src_len, enc, default_encoding, |
| 492 | &dst_len); |
| 493 | if (!dst) { |
| 494 | error(_("failed to encode '%s' from %s to %s"), |
| 495 | path, default_encoding, enc); |
| 496 | return 0; |
| 497 | } |
| 498 | |
| 499 | strbuf_attach(buf, dst, dst_len, dst_len + 1); |
| 500 | return 1; |
| 501 | } |
| 502 | |
| 503 | static int crlf_to_git(struct index_state *istate, |
| 504 | const char *path, const char *src, size_t len, |
| 505 | struct strbuf *buf, |
| 506 | enum convert_crlf_action crlf_action, int conv_flags) |
| 507 | { |
| 508 | struct text_stat stats; |
| 509 | char *dst; |
| 510 | int convert_crlf_into_lf; |
| 511 | |
| 512 | if (crlf_action == CRLF_BINARY || |
| 513 | (src && !len)) |
| 514 | return 0; |
| 515 | |
| 516 | /* |
| 517 | * If we are doing a dry-run and have no source buffer, there is |
| 518 | * nothing to analyze; we must assume we would convert. |
| 519 | */ |
| 520 | if (!buf && !src) |
| 521 | return 1; |
| 522 | |
| 523 | gather_stats(src, len, &stats); |
| 524 | /* Optimization: No CRLF? Nothing to convert, regardless. */ |
| 525 | convert_crlf_into_lf = !!stats.crlf; |
| 526 | |
| 527 | if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) { |
| 528 | if (convert_is_binary(&stats)) |
| 529 | return 0; |
| 530 | /* |
| 531 | * If the file in the index has any CR in it, do not |
| 532 | * convert. This is the new safer autocrlf handling, |
| 533 | * unless we want to renormalize in a merge or |
| 534 | * cherry-pick. |
| 535 | */ |
| 536 | if ((!(conv_flags & CONV_EOL_RENORMALIZE)) && |
| 537 | has_crlf_in_index(istate, path)) |
| 538 | convert_crlf_into_lf = 0; |
| 539 | } |
| 540 | if (((conv_flags & CONV_EOL_RNDTRP_WARN) || |
| 541 | ((conv_flags & CONV_EOL_RNDTRP_DIE) && len))) { |
| 542 | struct text_stat new_stats; |
| 543 | memcpy(&new_stats, &stats, sizeof(new_stats)); |
| 544 | /* simulate "git add" */ |
| 545 | if (convert_crlf_into_lf) { |
| 546 | new_stats.lonelf += new_stats.crlf; |
| 547 | new_stats.crlf = 0; |
| 548 | } |
| 549 | /* simulate "git checkout" */ |
| 550 | if (will_convert_lf_to_crlf(&new_stats, crlf_action)) { |
| 551 | new_stats.crlf += new_stats.lonelf; |
| 552 | new_stats.lonelf = 0; |
| 553 | } |
| 554 | check_global_conv_flags_eol(path, &stats, &new_stats, conv_flags); |
| 555 | } |
| 556 | if (!convert_crlf_into_lf) |
| 557 | return 0; |
| 558 | |
| 559 | /* |
| 560 | * At this point all of our source analysis is done, and we are sure we |
| 561 | * would convert. If we are in dry-run mode, we can give an answer. |
| 562 | */ |
| 563 | if (!buf) |
| 564 | return 1; |
| 565 | |
| 566 | /* only grow if not in place */ |
| 567 | if (strbuf_avail(buf) + buf->len < len) |
| 568 | strbuf_grow(buf, len - buf->len); |
| 569 | dst = buf->buf; |
| 570 | if (crlf_action == CRLF_AUTO || crlf_action == CRLF_AUTO_INPUT || crlf_action == CRLF_AUTO_CRLF) { |
| 571 | /* |
| 572 | * If we guessed, we already know we rejected a file with |
| 573 | * lone CR, and we can strip a CR without looking at what |
| 574 | * follow it. |
| 575 | */ |
| 576 | do { |
| 577 | unsigned char c = *src++; |
| 578 | if (c != '\r') |
| 579 | *dst++ = c; |
| 580 | } while (--len); |
| 581 | } else { |
| 582 | do { |
| 583 | unsigned char c = *src++; |
| 584 | if (! (c == '\r' && (1 < len && *src == '\n'))) |
| 585 | *dst++ = c; |
| 586 | } while (--len); |
| 587 | } |
| 588 | strbuf_setlen(buf, dst - buf->buf); |
| 589 | return 1; |
| 590 | } |
| 591 | |
| 592 | static int crlf_to_worktree(const char *src, size_t len, struct strbuf *buf, |
| 593 | enum convert_crlf_action crlf_action) |
| 594 | { |
| 595 | char *to_free = NULL; |
| 596 | struct text_stat stats; |
| 597 | |
| 598 | if (!len || output_eol(crlf_action) != EOL_CRLF) |
| 599 | return 0; |
| 600 | |
| 601 | gather_stats(src, len, &stats); |
| 602 | if (!will_convert_lf_to_crlf(&stats, crlf_action)) |
| 603 | return 0; |
| 604 | |
| 605 | /* are we "faking" in place editing ? */ |
| 606 | if (src == buf->buf) |
| 607 | to_free = strbuf_detach(buf, NULL); |
| 608 | |
| 609 | strbuf_grow(buf, len + stats.lonelf); |
| 610 | for (;;) { |
| 611 | const char *nl = memchr(src, '\n', len); |
| 612 | if (!nl) |
| 613 | break; |
| 614 | if (nl > src && nl[-1] == '\r') { |
| 615 | strbuf_add(buf, src, nl + 1 - src); |
| 616 | } else { |
| 617 | strbuf_add(buf, src, nl - src); |
| 618 | strbuf_addstr(buf, "\r\n"); |
| 619 | } |
| 620 | len -= nl + 1 - src; |
| 621 | src = nl + 1; |
| 622 | } |
| 623 | strbuf_add(buf, src, len); |
| 624 | |
| 625 | free(to_free); |
| 626 | return 1; |
| 627 | } |
| 628 | |
| 629 | struct filter_params { |
| 630 | const char *src; |
| 631 | size_t size; |
| 632 | int fd; |
| 633 | const char *cmd; |
| 634 | const char *path; |
| 635 | }; |
| 636 | |
| 637 | static int filter_buffer_or_fd(int in UNUSED, int out, void *data) |
| 638 | { |
| 639 | /* |
| 640 | * Spawn cmd and feed the buffer contents through its stdin. |
| 641 | */ |
| 642 | struct child_process child_process = CHILD_PROCESS_INIT; |
| 643 | struct filter_params *params = (struct filter_params *)data; |
| 644 | const char *format = params->cmd; |
| 645 | int write_err, status; |
| 646 | |
| 647 | /* apply % substitution to cmd */ |
| 648 | struct strbuf cmd = STRBUF_INIT; |
| 649 | |
| 650 | /* expand all %f with the quoted path; quote to preserve space, etc. */ |
| 651 | while (strbuf_expand_step(&cmd, &format)) { |
| 652 | if (skip_prefix(format, "%", &format)) |
| 653 | strbuf_addch(&cmd, '%'); |
| 654 | else if (skip_prefix(format, "f", &format)) |
| 655 | sq_quote_buf(&cmd, params->path); |
| 656 | else |
| 657 | strbuf_addch(&cmd, '%'); |
| 658 | } |
| 659 | |
| 660 | strvec_push(&child_process.args, cmd.buf); |
| 661 | child_process.use_shell = 1; |
| 662 | child_process.in = -1; |
| 663 | child_process.out = out; |
| 664 | |
| 665 | if (start_command(&child_process)) { |
| 666 | strbuf_release(&cmd); |
| 667 | return error(_("cannot fork to run external filter '%s'"), |
| 668 | params->cmd); |
| 669 | } |
| 670 | |
| 671 | sigchain_push(SIGPIPE, SIG_IGN); |
| 672 | |
| 673 | if (params->src) { |
| 674 | write_err = (write_in_full(child_process.in, |
| 675 | params->src, params->size) < 0); |
| 676 | if (errno == EPIPE) |
| 677 | write_err = 0; |
| 678 | } else { |
| 679 | write_err = copy_fd(params->fd, child_process.in); |
| 680 | if (write_err == COPY_WRITE_ERROR && errno == EPIPE) |
| 681 | write_err = 0; |
| 682 | } |
| 683 | |
| 684 | if (close(child_process.in)) |
| 685 | write_err = 1; |
| 686 | if (write_err) |
| 687 | error(_("cannot feed the input to external filter '%s'"), |
| 688 | params->cmd); |
| 689 | |
| 690 | sigchain_pop(SIGPIPE); |
| 691 | |
| 692 | status = finish_command(&child_process); |
| 693 | if (status) |
| 694 | error(_("external filter '%s' failed %d"), params->cmd, status); |
| 695 | |
| 696 | strbuf_release(&cmd); |
| 697 | return (write_err || status); |
| 698 | } |
| 699 | |
| 700 | static int apply_single_file_filter(const char *path, const char *src, size_t len, int fd, |
| 701 | struct strbuf *dst, const char *cmd) |
| 702 | { |
| 703 | /* |
| 704 | * Create a pipeline to have the command filter the buffer's |
| 705 | * contents. |
| 706 | * |
| 707 | * (child --> cmd) --> us |
| 708 | */ |
| 709 | int err = 0; |
| 710 | struct strbuf nbuf = STRBUF_INIT; |
| 711 | struct async async; |
| 712 | struct filter_params params; |
| 713 | |
| 714 | memset(&async, 0, sizeof(async)); |
| 715 | async.proc = filter_buffer_or_fd; |
| 716 | async.data = ¶ms; |
| 717 | async.out = -1; |
| 718 | params.src = src; |
| 719 | params.size = len; |
| 720 | params.fd = fd; |
| 721 | params.cmd = cmd; |
| 722 | params.path = path; |
| 723 | |
| 724 | fflush(NULL); |
| 725 | if (start_async(&async)) |
| 726 | return 0; /* error was already reported */ |
| 727 | |
| 728 | if (strbuf_read(&nbuf, async.out, 0) < 0) { |
| 729 | err = error(_("read from external filter '%s' failed"), cmd); |
| 730 | } |
| 731 | if (close(async.out)) { |
| 732 | err = error(_("read from external filter '%s' failed"), cmd); |
| 733 | } |
| 734 | if (finish_async(&async)) { |
| 735 | err = error(_("external filter '%s' failed"), cmd); |
| 736 | } |
| 737 | |
| 738 | if (!err) { |
| 739 | strbuf_swap(dst, &nbuf); |
| 740 | } |
| 741 | strbuf_release(&nbuf); |
| 742 | return !err; |
| 743 | } |
| 744 | |
| 745 | #define CAP_CLEAN (1u<<0) |
| 746 | #define CAP_SMUDGE (1u<<1) |
| 747 | #define CAP_DELAY (1u<<2) |
| 748 | |
| 749 | struct cmd2process { |
| 750 | struct subprocess_entry subprocess; /* must be the first member! */ |
| 751 | unsigned int supported_capabilities; |
| 752 | }; |
| 753 | |
| 754 | static int subprocess_map_initialized; |
| 755 | static struct hashmap subprocess_map; |
| 756 | |
| 757 | static int start_multi_file_filter_fn(struct subprocess_entry *subprocess) |
| 758 | { |
| 759 | static int versions[] = {2, 0}; |
| 760 | static struct subprocess_capability capabilities[] = { |
| 761 | { "clean", CAP_CLEAN }, |
| 762 | { "smudge", CAP_SMUDGE }, |
| 763 | { "delay", CAP_DELAY }, |
| 764 | { NULL, 0 } |
| 765 | }; |
| 766 | struct cmd2process *entry = (struct cmd2process *)subprocess; |
| 767 | return subprocess_handshake(subprocess, "git-filter", versions, NULL, |
| 768 | capabilities, |
| 769 | &entry->supported_capabilities); |
| 770 | } |
| 771 | |
| 772 | static void handle_filter_error(const struct strbuf *filter_status, |
| 773 | struct cmd2process *entry, |
| 774 | const unsigned int wanted_capability) |
| 775 | { |
| 776 | if (!strcmp(filter_status->buf, "error")) |
| 777 | ; /* The filter signaled a problem with the file. */ |
| 778 | else if (!strcmp(filter_status->buf, "abort") && wanted_capability) { |
| 779 | /* |
| 780 | * The filter signaled a permanent problem. Don't try to filter |
| 781 | * files with the same command for the lifetime of the current |
| 782 | * Git process. |
| 783 | */ |
| 784 | entry->supported_capabilities &= ~wanted_capability; |
| 785 | } else { |
| 786 | /* |
| 787 | * Something went wrong with the protocol filter. |
| 788 | * Force shutdown and restart if another blob requires filtering. |
| 789 | */ |
| 790 | error(_("external filter '%s' failed"), entry->subprocess.cmd); |
| 791 | subprocess_stop(&subprocess_map, &entry->subprocess); |
| 792 | free(entry); |
| 793 | } |
| 794 | } |
| 795 | |
| 796 | static int apply_multi_file_filter(const char *path, const char *src, size_t len, |
| 797 | int fd, struct strbuf *dst, const char *cmd, |
| 798 | const unsigned int wanted_capability, |
| 799 | const struct checkout_metadata *meta, |
| 800 | struct delayed_checkout *dco) |
| 801 | { |
| 802 | int err; |
| 803 | int can_delay = 0; |
| 804 | struct cmd2process *entry; |
| 805 | struct child_process *process; |
| 806 | struct strbuf nbuf = STRBUF_INIT; |
| 807 | struct strbuf filter_status = STRBUF_INIT; |
| 808 | const char *filter_type; |
| 809 | |
| 810 | if (!subprocess_map_initialized) { |
| 811 | subprocess_map_initialized = 1; |
| 812 | hashmap_init(&subprocess_map, cmd2process_cmp, NULL, 0); |
| 813 | entry = NULL; |
| 814 | } else { |
| 815 | entry = (struct cmd2process *)subprocess_find_entry(&subprocess_map, cmd); |
| 816 | } |
| 817 | |
| 818 | fflush(NULL); |
| 819 | |
| 820 | if (!entry) { |
| 821 | entry = xmalloc(sizeof(*entry)); |
| 822 | entry->supported_capabilities = 0; |
| 823 | |
| 824 | if (subprocess_start(&subprocess_map, &entry->subprocess, cmd, start_multi_file_filter_fn)) { |
| 825 | free(entry); |
| 826 | return 0; |
| 827 | } |
| 828 | } |
| 829 | process = &entry->subprocess.process; |
| 830 | |
| 831 | if (!(entry->supported_capabilities & wanted_capability)) |
| 832 | return 0; |
| 833 | |
| 834 | if (wanted_capability & CAP_CLEAN) |
| 835 | filter_type = "clean"; |
| 836 | else if (wanted_capability & CAP_SMUDGE) |
| 837 | filter_type = "smudge"; |
| 838 | else |
| 839 | die(_("unexpected filter type")); |
| 840 | |
| 841 | sigchain_push(SIGPIPE, SIG_IGN); |
| 842 | |
| 843 | assert(strlen(filter_type) < LARGE_PACKET_DATA_MAX - strlen("command=\n")); |
| 844 | err = packet_write_fmt_gently(process->in, "command=%s\n", filter_type); |
| 845 | if (err) |
| 846 | goto done; |
| 847 | |
| 848 | err = strlen(path) > LARGE_PACKET_DATA_MAX - strlen("pathname=\n"); |
| 849 | if (err) { |
| 850 | error(_("path name too long for external filter")); |
| 851 | goto done; |
| 852 | } |
| 853 | |
| 854 | err = packet_write_fmt_gently(process->in, "pathname=%s\n", path); |
| 855 | if (err) |
| 856 | goto done; |
| 857 | |
| 858 | if (meta && meta->refname) { |
| 859 | err = packet_write_fmt_gently(process->in, "ref=%s\n", meta->refname); |
| 860 | if (err) |
| 861 | goto done; |
| 862 | } |
| 863 | |
| 864 | if (meta && !is_null_oid(&meta->treeish)) { |
| 865 | err = packet_write_fmt_gently(process->in, "treeish=%s\n", oid_to_hex(&meta->treeish)); |
| 866 | if (err) |
| 867 | goto done; |
| 868 | } |
| 869 | |
| 870 | if (meta && !is_null_oid(&meta->blob)) { |
| 871 | err = packet_write_fmt_gently(process->in, "blob=%s\n", oid_to_hex(&meta->blob)); |
| 872 | if (err) |
| 873 | goto done; |
| 874 | } |
| 875 | |
| 876 | if ((entry->supported_capabilities & CAP_DELAY) && |
| 877 | dco && dco->state == CE_CAN_DELAY) { |
| 878 | can_delay = 1; |
| 879 | err = packet_write_fmt_gently(process->in, "can-delay=1\n"); |
| 880 | if (err) |
| 881 | goto done; |
| 882 | } |
| 883 | |
| 884 | err = packet_flush_gently(process->in); |
| 885 | if (err) |
| 886 | goto done; |
| 887 | |
| 888 | if (fd >= 0) |
| 889 | err = write_packetized_from_fd_no_flush(fd, process->in); |
| 890 | else |
| 891 | err = write_packetized_from_buf_no_flush(src, len, process->in); |
| 892 | if (err) |
| 893 | goto done; |
| 894 | |
| 895 | err = packet_flush_gently(process->in); |
| 896 | if (err) |
| 897 | goto done; |
| 898 | |
| 899 | err = subprocess_read_status(process->out, &filter_status); |
| 900 | if (err) |
| 901 | goto done; |
| 902 | |
| 903 | if (can_delay && !strcmp(filter_status.buf, "delayed")) { |
| 904 | string_list_insert(&dco->filters, cmd); |
| 905 | string_list_insert(&dco->paths, path); |
| 906 | } else { |
| 907 | /* The filter got the blob and wants to send us a response. */ |
| 908 | err = strcmp(filter_status.buf, "success"); |
| 909 | if (err) |
| 910 | goto done; |
| 911 | |
| 912 | err = read_packetized_to_strbuf(process->out, &nbuf, |
| 913 | PACKET_READ_GENTLE_ON_EOF) < 0; |
| 914 | if (err) |
| 915 | goto done; |
| 916 | |
| 917 | err = subprocess_read_status(process->out, &filter_status); |
| 918 | if (err) |
| 919 | goto done; |
| 920 | |
| 921 | err = strcmp(filter_status.buf, "success"); |
| 922 | } |
| 923 | |
| 924 | done: |
| 925 | sigchain_pop(SIGPIPE); |
| 926 | |
| 927 | if (err) |
| 928 | handle_filter_error(&filter_status, entry, wanted_capability); |
| 929 | else |
| 930 | strbuf_swap(dst, &nbuf); |
| 931 | strbuf_release(&nbuf); |
| 932 | strbuf_release(&filter_status); |
| 933 | return !err; |
| 934 | } |
| 935 | |
| 936 | |
| 937 | int async_query_available_blobs(const char *cmd, struct string_list *available_paths) |
| 938 | { |
| 939 | int err; |
| 940 | char *line; |
| 941 | struct cmd2process *entry; |
| 942 | struct child_process *process; |
| 943 | struct strbuf filter_status = STRBUF_INIT; |
| 944 | |
| 945 | assert(subprocess_map_initialized); |
| 946 | entry = (struct cmd2process *)subprocess_find_entry(&subprocess_map, cmd); |
| 947 | if (!entry) { |
| 948 | error(_("external filter '%s' is not available anymore although " |
| 949 | "not all paths have been filtered"), cmd); |
| 950 | return 0; |
| 951 | } |
| 952 | process = &entry->subprocess.process; |
| 953 | sigchain_push(SIGPIPE, SIG_IGN); |
| 954 | |
| 955 | err = packet_write_fmt_gently( |
| 956 | process->in, "command=list_available_blobs\n"); |
| 957 | if (err) |
| 958 | goto done; |
| 959 | |
| 960 | err = packet_flush_gently(process->in); |
| 961 | if (err) |
| 962 | goto done; |
| 963 | |
| 964 | while ((line = packet_read_line(process->out, NULL))) { |
| 965 | const char *path; |
| 966 | if (skip_prefix(line, "pathname=", &path)) |
| 967 | string_list_insert(available_paths, path); |
| 968 | else |
| 969 | ; /* ignore unknown keys */ |
| 970 | } |
| 971 | |
| 972 | err = subprocess_read_status(process->out, &filter_status); |
| 973 | if (err) |
| 974 | goto done; |
| 975 | |
| 976 | err = strcmp(filter_status.buf, "success"); |
| 977 | |
| 978 | done: |
| 979 | sigchain_pop(SIGPIPE); |
| 980 | |
| 981 | if (err) |
| 982 | handle_filter_error(&filter_status, entry, 0); |
| 983 | strbuf_release(&filter_status); |
| 984 | return !err; |
| 985 | } |
| 986 | |
| 987 | static struct convert_driver { |
| 988 | const char *name; |
| 989 | struct convert_driver *next; |
| 990 | char *smudge; |
| 991 | char *clean; |
| 992 | char *process; |
| 993 | int required; |
| 994 | } *user_convert, **user_convert_tail; |
| 995 | |
| 996 | static int apply_filter(const char *path, const char *src, size_t len, |
| 997 | int fd, struct strbuf *dst, struct convert_driver *drv, |
| 998 | const unsigned int wanted_capability, |
| 999 | const struct checkout_metadata *meta, |
| 1000 | struct delayed_checkout *dco) |
| 1001 | { |
| 1002 | const char *cmd = NULL; |
| 1003 | |
| 1004 | if (!drv) |
| 1005 | return 0; |
| 1006 | |
| 1007 | if (!dst) |
| 1008 | return 1; |
| 1009 | |
| 1010 | if ((wanted_capability & CAP_CLEAN) && !drv->process && drv->clean) |
| 1011 | cmd = drv->clean; |
| 1012 | else if ((wanted_capability & CAP_SMUDGE) && !drv->process && drv->smudge) |
| 1013 | cmd = drv->smudge; |
| 1014 | |
| 1015 | if (cmd && *cmd) |
| 1016 | return apply_single_file_filter(path, src, len, fd, dst, cmd); |
| 1017 | else if (drv->process && *drv->process) |
| 1018 | return apply_multi_file_filter(path, src, len, fd, dst, |
| 1019 | drv->process, wanted_capability, meta, dco); |
| 1020 | |
| 1021 | return 0; |
| 1022 | } |
| 1023 | |
| 1024 | static int read_convert_config(const char *var, const char *value, |
| 1025 | const struct config_context *ctx UNUSED, |
| 1026 | void *cb UNUSED) |
| 1027 | { |
| 1028 | const char *key, *name; |
| 1029 | size_t namelen; |
| 1030 | struct convert_driver *drv; |
| 1031 | |
| 1032 | /* |
| 1033 | * External conversion drivers are configured using |
| 1034 | * "filter.<name>.variable". |
| 1035 | */ |
| 1036 | if (parse_config_key(var, "filter", &name, &namelen, &key) < 0 || !name) |
| 1037 | return 0; |
| 1038 | for (drv = user_convert; drv; drv = drv->next) |
| 1039 | if (!xstrncmpz(drv->name, name, namelen)) |
| 1040 | break; |
| 1041 | if (!drv) { |
| 1042 | CALLOC_ARRAY(drv, 1); |
| 1043 | drv->name = xmemdupz(name, namelen); |
| 1044 | *user_convert_tail = drv; |
| 1045 | user_convert_tail = &(drv->next); |
| 1046 | } |
| 1047 | |
| 1048 | /* |
| 1049 | * filter.<name>.smudge and filter.<name>.clean specifies |
| 1050 | * the command line: |
| 1051 | * |
| 1052 | * command-line |
| 1053 | * |
| 1054 | * The command-line will not be interpolated in any way. |
| 1055 | */ |
| 1056 | |
| 1057 | if (!strcmp("smudge", key)) { |
| 1058 | FREE_AND_NULL(drv->smudge); |
| 1059 | return git_config_string(&drv->smudge, var, value); |
| 1060 | } |
| 1061 | |
| 1062 | if (!strcmp("clean", key)) { |
| 1063 | FREE_AND_NULL(drv->clean); |
| 1064 | return git_config_string(&drv->clean, var, value); |
| 1065 | } |
| 1066 | |
| 1067 | if (!strcmp("process", key)) { |
| 1068 | FREE_AND_NULL(drv->process); |
| 1069 | return git_config_string(&drv->process, var, value); |
| 1070 | } |
| 1071 | |
| 1072 | if (!strcmp("required", key)) { |
| 1073 | drv->required = git_config_bool(var, value); |
| 1074 | return 0; |
| 1075 | } |
| 1076 | |
| 1077 | return 0; |
| 1078 | } |
| 1079 | |
| 1080 | static int count_ident(const char *cp, unsigned long size) |
| 1081 | { |
| 1082 | /* |
| 1083 | * "$Id: 0000000000000000000000000000000000000000 $" <=> "$Id$" |
| 1084 | */ |
| 1085 | int cnt = 0; |
| 1086 | char ch; |
| 1087 | |
| 1088 | while (size) { |
| 1089 | ch = *cp++; |
| 1090 | size--; |
| 1091 | if (ch != '$') |
| 1092 | continue; |
| 1093 | if (size < 3) |
| 1094 | break; |
| 1095 | if (memcmp("Id", cp, 2)) |
| 1096 | continue; |
| 1097 | ch = cp[2]; |
| 1098 | cp += 3; |
| 1099 | size -= 3; |
| 1100 | if (ch == '$') |
| 1101 | cnt++; /* $Id$ */ |
| 1102 | if (ch != ':') |
| 1103 | continue; |
| 1104 | |
| 1105 | /* |
| 1106 | * "$Id: ... "; scan up to the closing dollar sign and discard. |
| 1107 | */ |
| 1108 | while (size) { |
| 1109 | ch = *cp++; |
| 1110 | size--; |
| 1111 | if (ch == '$') { |
| 1112 | cnt++; |
| 1113 | break; |
| 1114 | } |
| 1115 | if (ch == '\n') |
| 1116 | break; |
| 1117 | } |
| 1118 | } |
| 1119 | return cnt; |
| 1120 | } |
| 1121 | |
| 1122 | static int ident_to_git(const char *src, size_t len, |
| 1123 | struct strbuf *buf, int ident) |
| 1124 | { |
| 1125 | char *dst; |
| 1126 | const char *dollar; |
| 1127 | |
| 1128 | if (!ident || (src && !count_ident(src, len))) |
| 1129 | return 0; |
| 1130 | |
| 1131 | if (!buf) |
| 1132 | return 1; |
| 1133 | |
| 1134 | /* only grow if not in place */ |
| 1135 | if (strbuf_avail(buf) + buf->len < len) |
| 1136 | strbuf_grow(buf, len - buf->len); |
| 1137 | dst = buf->buf; |
| 1138 | for (;;) { |
| 1139 | dollar = memchr(src, '$', len); |
| 1140 | if (!dollar) |
| 1141 | break; |
| 1142 | memmove(dst, src, dollar + 1 - src); |
| 1143 | dst += dollar + 1 - src; |
| 1144 | len -= dollar + 1 - src; |
| 1145 | src = dollar + 1; |
| 1146 | |
| 1147 | if (len > 3 && !memcmp(src, "Id:", 3)) { |
| 1148 | dollar = memchr(src + 3, '$', len - 3); |
| 1149 | if (!dollar) |
| 1150 | break; |
| 1151 | if (memchr(src + 3, '\n', dollar - src - 3)) { |
| 1152 | /* Line break before the next dollar. */ |
| 1153 | continue; |
| 1154 | } |
| 1155 | |
| 1156 | memcpy(dst, "Id$", 3); |
| 1157 | dst += 3; |
| 1158 | len -= dollar + 1 - src; |
| 1159 | src = dollar + 1; |
| 1160 | } |
| 1161 | } |
| 1162 | memmove(dst, src, len); |
| 1163 | strbuf_setlen(buf, dst + len - buf->buf); |
| 1164 | return 1; |
| 1165 | } |
| 1166 | |
| 1167 | static int ident_to_worktree(const char *src, size_t len, |
| 1168 | struct strbuf *buf, int ident) |
| 1169 | { |
| 1170 | struct object_id oid; |
| 1171 | char *to_free = NULL; |
| 1172 | const char *dollar, *spc; |
| 1173 | int cnt; |
| 1174 | |
| 1175 | if (!ident) |
| 1176 | return 0; |
| 1177 | |
| 1178 | cnt = count_ident(src, len); |
| 1179 | if (!cnt) |
| 1180 | return 0; |
| 1181 | |
| 1182 | /* are we "faking" in place editing ? */ |
| 1183 | if (src == buf->buf) |
| 1184 | to_free = strbuf_detach(buf, NULL); |
| 1185 | hash_object_file(the_hash_algo, src, len, OBJ_BLOB, &oid); |
| 1186 | |
| 1187 | strbuf_grow(buf, len + cnt * (the_hash_algo->hexsz + 3)); |
| 1188 | for (;;) { |
| 1189 | /* step 1: run to the next '$' */ |
| 1190 | dollar = memchr(src, '$', len); |
| 1191 | if (!dollar) |
| 1192 | break; |
| 1193 | strbuf_add(buf, src, dollar + 1 - src); |
| 1194 | len -= dollar + 1 - src; |
| 1195 | src = dollar + 1; |
| 1196 | |
| 1197 | /* step 2: does it looks like a bit like Id:xxx$ or Id$ ? */ |
| 1198 | if (len < 3 || memcmp("Id", src, 2)) |
| 1199 | continue; |
| 1200 | |
| 1201 | /* step 3: skip over Id$ or Id:xxxxx$ */ |
| 1202 | if (src[2] == '$') { |
| 1203 | src += 3; |
| 1204 | len -= 3; |
| 1205 | } else if (src[2] == ':') { |
| 1206 | /* |
| 1207 | * It's possible that an expanded Id has crept its way into the |
| 1208 | * repository, we cope with that by stripping the expansion out. |
| 1209 | * This is probably not a good idea, since it will cause changes |
| 1210 | * on checkout, which won't go away by stash, but let's keep it |
| 1211 | * for git-style ids. |
| 1212 | */ |
| 1213 | dollar = memchr(src + 3, '$', len - 3); |
| 1214 | if (!dollar) { |
| 1215 | /* incomplete keyword, no more '$', so just quit the loop */ |
| 1216 | break; |
| 1217 | } |
| 1218 | |
| 1219 | if (memchr(src + 3, '\n', dollar - src - 3)) { |
| 1220 | /* Line break before the next dollar. */ |
| 1221 | continue; |
| 1222 | } |
| 1223 | |
| 1224 | spc = memchr(src + 4, ' ', dollar - src - 4); |
| 1225 | if (spc && spc < dollar-1) { |
| 1226 | /* There are spaces in unexpected places. |
| 1227 | * This is probably an id from some other |
| 1228 | * versioning system. Keep it for now. |
| 1229 | */ |
| 1230 | continue; |
| 1231 | } |
| 1232 | |
| 1233 | len -= dollar + 1 - src; |
| 1234 | src = dollar + 1; |
| 1235 | } else { |
| 1236 | /* it wasn't a "Id$" or "Id:xxxx$" */ |
| 1237 | continue; |
| 1238 | } |
| 1239 | |
| 1240 | /* step 4: substitute */ |
| 1241 | strbuf_addstr(buf, "Id: "); |
| 1242 | strbuf_add_oid_hex(buf, &oid); |
| 1243 | strbuf_addstr(buf, " $"); |
| 1244 | } |
| 1245 | strbuf_add(buf, src, len); |
| 1246 | |
| 1247 | free(to_free); |
| 1248 | return 1; |
| 1249 | } |
| 1250 | |
| 1251 | static const char *git_path_check_encoding(struct attr_check_item *check) |
| 1252 | { |
| 1253 | const char *value = check->value; |
| 1254 | |
| 1255 | if (ATTR_UNSET(value) || !strlen(value)) |
| 1256 | return NULL; |
| 1257 | |
| 1258 | if (ATTR_TRUE(value) || ATTR_FALSE(value)) { |
| 1259 | die(_("true/false are no valid working-tree-encodings")); |
| 1260 | } |
| 1261 | |
| 1262 | /* Don't encode to the default encoding */ |
| 1263 | if (same_encoding(value, default_encoding)) |
| 1264 | return NULL; |
| 1265 | |
| 1266 | return value; |
| 1267 | } |
| 1268 | |
| 1269 | static enum convert_crlf_action git_path_check_crlf(struct attr_check_item *check) |
| 1270 | { |
| 1271 | const char *value = check->value; |
| 1272 | |
| 1273 | if (ATTR_TRUE(value)) |
| 1274 | return CRLF_TEXT; |
| 1275 | else if (ATTR_FALSE(value)) |
| 1276 | return CRLF_BINARY; |
| 1277 | else if (ATTR_UNSET(value)) |
| 1278 | ; |
| 1279 | else if (!strcmp(value, "input")) |
| 1280 | return CRLF_TEXT_INPUT; |
| 1281 | else if (!strcmp(value, "auto")) |
| 1282 | return CRLF_AUTO; |
| 1283 | return CRLF_UNDEFINED; |
| 1284 | } |
| 1285 | |
| 1286 | static enum eol git_path_check_eol(struct attr_check_item *check) |
| 1287 | { |
| 1288 | const char *value = check->value; |
| 1289 | |
| 1290 | if (ATTR_UNSET(value)) |
| 1291 | ; |
| 1292 | else if (!strcmp(value, "lf")) |
| 1293 | return EOL_LF; |
| 1294 | else if (!strcmp(value, "crlf")) |
| 1295 | return EOL_CRLF; |
| 1296 | return EOL_UNSET; |
| 1297 | } |
| 1298 | |
| 1299 | static struct convert_driver *git_path_check_convert(struct attr_check_item *check) |
| 1300 | { |
| 1301 | const char *value = check->value; |
| 1302 | struct convert_driver *drv; |
| 1303 | |
| 1304 | if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value)) |
| 1305 | return NULL; |
| 1306 | for (drv = user_convert; drv; drv = drv->next) |
| 1307 | if (!strcmp(value, drv->name)) |
| 1308 | return drv; |
| 1309 | return NULL; |
| 1310 | } |
| 1311 | |
| 1312 | static int git_path_check_ident(struct attr_check_item *check) |
| 1313 | { |
| 1314 | const char *value = check->value; |
| 1315 | |
| 1316 | return !!ATTR_TRUE(value); |
| 1317 | } |
| 1318 | |
| 1319 | static struct attr_check *check; |
| 1320 | |
| 1321 | void convert_attrs(struct index_state *istate, |
| 1322 | struct conv_attrs *ca, const char *path) |
| 1323 | { |
| 1324 | struct attr_check_item *ccheck = NULL; |
| 1325 | |
| 1326 | if (!check) { |
| 1327 | check = attr_check_initl("crlf", "ident", "filter", |
| 1328 | "eol", "text", "working-tree-encoding", |
| 1329 | NULL); |
| 1330 | user_convert_tail = &user_convert; |
| 1331 | repo_config(the_repository, read_convert_config, NULL); |
| 1332 | } |
| 1333 | |
| 1334 | git_check_attr(istate, path, check); |
| 1335 | ccheck = check->items; |
| 1336 | ca->crlf_action = git_path_check_crlf(ccheck + 4); |
| 1337 | if (ca->crlf_action == CRLF_UNDEFINED) |
| 1338 | ca->crlf_action = git_path_check_crlf(ccheck + 0); |
| 1339 | ca->ident = git_path_check_ident(ccheck + 1); |
| 1340 | ca->drv = git_path_check_convert(ccheck + 2); |
| 1341 | if (ca->crlf_action != CRLF_BINARY) { |
| 1342 | enum eol eol_attr = git_path_check_eol(ccheck + 3); |
| 1343 | if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_LF) |
| 1344 | ca->crlf_action = CRLF_AUTO_INPUT; |
| 1345 | else if (ca->crlf_action == CRLF_AUTO && eol_attr == EOL_CRLF) |
| 1346 | ca->crlf_action = CRLF_AUTO_CRLF; |
| 1347 | else if (eol_attr == EOL_LF) |
| 1348 | ca->crlf_action = CRLF_TEXT_INPUT; |
| 1349 | else if (eol_attr == EOL_CRLF) |
| 1350 | ca->crlf_action = CRLF_TEXT_CRLF; |
| 1351 | } |
| 1352 | ca->working_tree_encoding = git_path_check_encoding(ccheck + 5); |
| 1353 | |
| 1354 | /* Save attr and make a decision for action */ |
| 1355 | ca->attr_action = ca->crlf_action; |
| 1356 | if (ca->crlf_action == CRLF_TEXT) |
| 1357 | ca->crlf_action = text_eol_is_crlf() ? CRLF_TEXT_CRLF : CRLF_TEXT_INPUT; |
| 1358 | if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_FALSE) |
| 1359 | ca->crlf_action = CRLF_BINARY; |
| 1360 | if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_TRUE) |
| 1361 | ca->crlf_action = CRLF_AUTO_CRLF; |
| 1362 | if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_INPUT) |
| 1363 | ca->crlf_action = CRLF_AUTO_INPUT; |
| 1364 | } |
| 1365 | |
| 1366 | void reset_parsed_attributes(void) |
| 1367 | { |
| 1368 | struct convert_driver *drv, *next; |
| 1369 | |
| 1370 | attr_check_free(check); |
| 1371 | check = NULL; |
| 1372 | reset_merge_attributes(); |
| 1373 | |
| 1374 | for (drv = user_convert; drv; drv = next) { |
| 1375 | next = drv->next; |
| 1376 | free((void *)drv->name); |
| 1377 | free((void *)drv->smudge); |
| 1378 | free((void *)drv->clean); |
| 1379 | free((void *)drv->process); |
| 1380 | free(drv); |
| 1381 | } |
| 1382 | user_convert = NULL; |
| 1383 | user_convert_tail = NULL; |
| 1384 | } |
| 1385 | |
| 1386 | int would_convert_to_git_filter_fd(struct index_state *istate, const char *path) |
| 1387 | { |
| 1388 | struct conv_attrs ca; |
| 1389 | |
| 1390 | convert_attrs(istate, &ca, path); |
| 1391 | if (!ca.drv) |
| 1392 | return 0; |
| 1393 | |
| 1394 | /* |
| 1395 | * Apply a filter to an fd only if the filter is required to succeed. |
| 1396 | * We must die if the filter fails, because the original data before |
| 1397 | * filtering is not available. |
| 1398 | */ |
| 1399 | if (!ca.drv->required) |
| 1400 | return 0; |
| 1401 | |
| 1402 | return apply_filter(path, NULL, 0, -1, NULL, ca.drv, CAP_CLEAN, NULL, NULL); |
| 1403 | } |
| 1404 | |
| 1405 | const char *get_convert_attr_ascii(struct index_state *istate, const char *path) |
| 1406 | { |
| 1407 | struct conv_attrs ca; |
| 1408 | |
| 1409 | convert_attrs(istate, &ca, path); |
| 1410 | switch (ca.attr_action) { |
| 1411 | case CRLF_UNDEFINED: |
| 1412 | return ""; |
| 1413 | case CRLF_BINARY: |
| 1414 | return "-text"; |
| 1415 | case CRLF_TEXT: |
| 1416 | return "text"; |
| 1417 | case CRLF_TEXT_INPUT: |
| 1418 | return "text eol=lf"; |
| 1419 | case CRLF_TEXT_CRLF: |
| 1420 | return "text eol=crlf"; |
| 1421 | case CRLF_AUTO: |
| 1422 | return "text=auto"; |
| 1423 | case CRLF_AUTO_CRLF: |
| 1424 | return "text=auto eol=crlf"; |
| 1425 | case CRLF_AUTO_INPUT: |
| 1426 | return "text=auto eol=lf"; |
| 1427 | } |
| 1428 | return ""; |
| 1429 | } |
| 1430 | |
| 1431 | int convert_to_git(struct index_state *istate, |
| 1432 | const char *path, const char *src, size_t len, |
| 1433 | struct strbuf *dst, int conv_flags) |
| 1434 | { |
| 1435 | int ret = 0; |
| 1436 | struct conv_attrs ca; |
| 1437 | |
| 1438 | convert_attrs(istate, &ca, path); |
| 1439 | |
| 1440 | ret |= apply_filter(path, src, len, -1, dst, ca.drv, CAP_CLEAN, NULL, NULL); |
| 1441 | if (!ret && ca.drv && ca.drv->required) |
| 1442 | die(_("%s: clean filter '%s' failed"), path, ca.drv->name); |
| 1443 | |
| 1444 | if (ret && dst) { |
| 1445 | src = dst->buf; |
| 1446 | len = dst->len; |
| 1447 | } |
| 1448 | |
| 1449 | ret |= encode_to_git(path, src, len, dst, ca.working_tree_encoding, conv_flags); |
| 1450 | if (ret && dst) { |
| 1451 | src = dst->buf; |
| 1452 | len = dst->len; |
| 1453 | } |
| 1454 | |
| 1455 | if (!(conv_flags & CONV_EOL_KEEP_CRLF)) { |
| 1456 | ret |= crlf_to_git(istate, path, src, len, dst, ca.crlf_action, conv_flags); |
| 1457 | if (ret && dst) { |
| 1458 | src = dst->buf; |
| 1459 | len = dst->len; |
| 1460 | } |
| 1461 | } |
| 1462 | return ret | ident_to_git(src, len, dst, ca.ident); |
| 1463 | } |
| 1464 | |
| 1465 | void convert_to_git_filter_fd(struct index_state *istate, |
| 1466 | const char *path, int fd, struct strbuf *dst, |
| 1467 | int conv_flags) |
| 1468 | { |
| 1469 | struct conv_attrs ca; |
| 1470 | convert_attrs(istate, &ca, path); |
| 1471 | |
| 1472 | assert(ca.drv); |
| 1473 | |
| 1474 | if (!apply_filter(path, NULL, 0, fd, dst, ca.drv, CAP_CLEAN, NULL, NULL)) |
| 1475 | die(_("%s: clean filter '%s' failed"), path, ca.drv->name); |
| 1476 | |
| 1477 | encode_to_git(path, dst->buf, dst->len, dst, ca.working_tree_encoding, conv_flags); |
| 1478 | crlf_to_git(istate, path, dst->buf, dst->len, dst, ca.crlf_action, conv_flags); |
| 1479 | ident_to_git(dst->buf, dst->len, dst, ca.ident); |
| 1480 | } |
| 1481 | |
| 1482 | static int convert_to_working_tree_ca_internal(const struct conv_attrs *ca, |
| 1483 | const char *path, const char *src, |
| 1484 | size_t len, struct strbuf *dst, |
| 1485 | int normalizing, |
| 1486 | const struct checkout_metadata *meta, |
| 1487 | struct delayed_checkout *dco) |
| 1488 | { |
| 1489 | int ret = 0, ret_filter = 0; |
| 1490 | |
| 1491 | ret |= ident_to_worktree(src, len, dst, ca->ident); |
| 1492 | if (ret) { |
| 1493 | src = dst->buf; |
| 1494 | len = dst->len; |
| 1495 | } |
| 1496 | /* |
| 1497 | * CRLF conversion can be skipped if normalizing, unless there |
| 1498 | * is a smudge or process filter (even if the process filter doesn't |
| 1499 | * support smudge). The filters might expect CRLFs. |
| 1500 | */ |
| 1501 | if ((ca->drv && (ca->drv->smudge || ca->drv->process)) || !normalizing) { |
| 1502 | ret |= crlf_to_worktree(src, len, dst, ca->crlf_action); |
| 1503 | if (ret) { |
| 1504 | src = dst->buf; |
| 1505 | len = dst->len; |
| 1506 | } |
| 1507 | } |
| 1508 | |
| 1509 | ret |= encode_to_worktree(path, src, len, dst, ca->working_tree_encoding); |
| 1510 | if (ret) { |
| 1511 | src = dst->buf; |
| 1512 | len = dst->len; |
| 1513 | } |
| 1514 | |
| 1515 | ret_filter = apply_filter( |
| 1516 | path, src, len, -1, dst, ca->drv, CAP_SMUDGE, meta, dco); |
| 1517 | if (!ret_filter && ca->drv && ca->drv->required) |
| 1518 | die(_("%s: smudge filter %s failed"), path, ca->drv->name); |
| 1519 | |
| 1520 | return ret | ret_filter; |
| 1521 | } |
| 1522 | |
| 1523 | int async_convert_to_working_tree_ca(const struct conv_attrs *ca, |
| 1524 | const char *path, const char *src, |
| 1525 | size_t len, struct strbuf *dst, |
| 1526 | const struct checkout_metadata *meta, |
| 1527 | void *dco) |
| 1528 | { |
| 1529 | return convert_to_working_tree_ca_internal(ca, path, src, len, dst, 0, |
| 1530 | meta, dco); |
| 1531 | } |
| 1532 | |
| 1533 | int convert_to_working_tree_ca(const struct conv_attrs *ca, |
| 1534 | const char *path, const char *src, |
| 1535 | size_t len, struct strbuf *dst, |
| 1536 | const struct checkout_metadata *meta) |
| 1537 | { |
| 1538 | return convert_to_working_tree_ca_internal(ca, path, src, len, dst, 0, |
| 1539 | meta, NULL); |
| 1540 | } |
| 1541 | |
| 1542 | int renormalize_buffer(struct index_state *istate, const char *path, |
| 1543 | const char *src, size_t len, struct strbuf *dst) |
| 1544 | { |
| 1545 | struct conv_attrs ca; |
| 1546 | int ret; |
| 1547 | |
| 1548 | convert_attrs(istate, &ca, path); |
| 1549 | ret = convert_to_working_tree_ca_internal(&ca, path, src, len, dst, 1, |
| 1550 | NULL, NULL); |
| 1551 | if (ret) { |
| 1552 | src = dst->buf; |
| 1553 | len = dst->len; |
| 1554 | } |
| 1555 | return ret | convert_to_git(istate, path, src, len, dst, CONV_EOL_RENORMALIZE); |
| 1556 | } |
| 1557 | |
| 1558 | /***************************************************************** |
| 1559 | * |
| 1560 | * Streaming conversion support |
| 1561 | * |
| 1562 | *****************************************************************/ |
| 1563 | |
| 1564 | typedef int (*filter_fn)(struct stream_filter *, |
| 1565 | const char *input, size_t *isize_p, |
| 1566 | char *output, size_t *osize_p); |
| 1567 | typedef void (*free_fn)(struct stream_filter *); |
| 1568 | |
| 1569 | struct stream_filter_vtbl { |
| 1570 | filter_fn filter; |
| 1571 | free_fn free; |
| 1572 | }; |
| 1573 | |
| 1574 | struct stream_filter { |
| 1575 | struct stream_filter_vtbl *vtbl; |
| 1576 | }; |
| 1577 | |
| 1578 | static int null_filter_fn(struct stream_filter *filter UNUSED, |
| 1579 | const char *input, size_t *isize_p, |
| 1580 | char *output, size_t *osize_p) |
| 1581 | { |
| 1582 | size_t count; |
| 1583 | |
| 1584 | if (!input) |
| 1585 | return 0; /* we do not keep any states */ |
| 1586 | count = *isize_p; |
| 1587 | if (*osize_p < count) |
| 1588 | count = *osize_p; |
| 1589 | if (count) { |
| 1590 | memmove(output, input, count); |
| 1591 | *isize_p -= count; |
| 1592 | *osize_p -= count; |
| 1593 | } |
| 1594 | return 0; |
| 1595 | } |
| 1596 | |
| 1597 | static void null_free_fn(struct stream_filter *filter UNUSED) |
| 1598 | { |
| 1599 | ; /* nothing -- null instances are shared */ |
| 1600 | } |
| 1601 | |
| 1602 | static struct stream_filter_vtbl null_vtbl = { |
| 1603 | .filter = null_filter_fn, |
| 1604 | .free = null_free_fn, |
| 1605 | }; |
| 1606 | |
| 1607 | static struct stream_filter null_filter_singleton = { |
| 1608 | .vtbl = &null_vtbl, |
| 1609 | }; |
| 1610 | |
| 1611 | int is_null_stream_filter(struct stream_filter *filter) |
| 1612 | { |
| 1613 | return filter == &null_filter_singleton; |
| 1614 | } |
| 1615 | |
| 1616 | |
| 1617 | /* |
| 1618 | * LF-to-CRLF filter |
| 1619 | */ |
| 1620 | |
| 1621 | struct lf_to_crlf_filter { |
| 1622 | struct stream_filter filter; |
| 1623 | unsigned has_held:1; |
| 1624 | char held; |
| 1625 | }; |
| 1626 | |
| 1627 | static int lf_to_crlf_filter_fn(struct stream_filter *filter, |
| 1628 | const char *input, size_t *isize_p, |
| 1629 | char *output, size_t *osize_p) |
| 1630 | { |
| 1631 | size_t count, o = 0; |
| 1632 | struct lf_to_crlf_filter *lf_to_crlf = (struct lf_to_crlf_filter *)filter; |
| 1633 | |
| 1634 | /* |
| 1635 | * We may be holding onto the CR to see if it is followed by a |
| 1636 | * LF, in which case we would need to go to the main loop. |
| 1637 | * Otherwise, just emit it to the output stream. |
| 1638 | */ |
| 1639 | if (lf_to_crlf->has_held && (lf_to_crlf->held != '\r' || !input)) { |
| 1640 | output[o++] = lf_to_crlf->held; |
| 1641 | lf_to_crlf->has_held = 0; |
| 1642 | } |
| 1643 | |
| 1644 | /* We are told to drain */ |
| 1645 | if (!input) { |
| 1646 | *osize_p -= o; |
| 1647 | return 0; |
| 1648 | } |
| 1649 | |
| 1650 | count = *isize_p; |
| 1651 | if (count || lf_to_crlf->has_held) { |
| 1652 | size_t i; |
| 1653 | int was_cr = 0; |
| 1654 | |
| 1655 | if (lf_to_crlf->has_held) { |
| 1656 | was_cr = 1; |
| 1657 | lf_to_crlf->has_held = 0; |
| 1658 | } |
| 1659 | |
| 1660 | for (i = 0; o < *osize_p && i < count; i++) { |
| 1661 | char ch = input[i]; |
| 1662 | |
| 1663 | if (ch == '\n') { |
| 1664 | output[o++] = '\r'; |
| 1665 | } else if (was_cr) { |
| 1666 | /* |
| 1667 | * Previous round saw CR and it is not followed |
| 1668 | * by a LF; emit the CR before processing the |
| 1669 | * current character. |
| 1670 | */ |
| 1671 | output[o++] = '\r'; |
| 1672 | } |
| 1673 | |
| 1674 | /* |
| 1675 | * We may have consumed the last output slot, |
| 1676 | * in which case we need to break out of this |
| 1677 | * loop; hold the current character before |
| 1678 | * returning. |
| 1679 | */ |
| 1680 | if (*osize_p <= o) { |
| 1681 | lf_to_crlf->has_held = 1; |
| 1682 | lf_to_crlf->held = ch; |
| 1683 | continue; /* break but increment i */ |
| 1684 | } |
| 1685 | |
| 1686 | if (ch == '\r') { |
| 1687 | was_cr = 1; |
| 1688 | continue; |
| 1689 | } |
| 1690 | |
| 1691 | was_cr = 0; |
| 1692 | output[o++] = ch; |
| 1693 | } |
| 1694 | |
| 1695 | *osize_p -= o; |
| 1696 | *isize_p -= i; |
| 1697 | |
| 1698 | if (!lf_to_crlf->has_held && was_cr) { |
| 1699 | lf_to_crlf->has_held = 1; |
| 1700 | lf_to_crlf->held = '\r'; |
| 1701 | } |
| 1702 | } |
| 1703 | return 0; |
| 1704 | } |
| 1705 | |
| 1706 | static void lf_to_crlf_free_fn(struct stream_filter *filter) |
| 1707 | { |
| 1708 | free(filter); |
| 1709 | } |
| 1710 | |
| 1711 | static struct stream_filter_vtbl lf_to_crlf_vtbl = { |
| 1712 | .filter = lf_to_crlf_filter_fn, |
| 1713 | .free = lf_to_crlf_free_fn, |
| 1714 | }; |
| 1715 | |
| 1716 | static struct stream_filter *lf_to_crlf_filter(void) |
| 1717 | { |
| 1718 | struct lf_to_crlf_filter *lf_to_crlf = xcalloc(1, sizeof(*lf_to_crlf)); |
| 1719 | |
| 1720 | lf_to_crlf->filter.vtbl = &lf_to_crlf_vtbl; |
| 1721 | return (struct stream_filter *)lf_to_crlf; |
| 1722 | } |
| 1723 | |
| 1724 | /* |
| 1725 | * Cascade filter |
| 1726 | */ |
| 1727 | #define FILTER_BUFFER 1024 |
| 1728 | struct cascade_filter { |
| 1729 | struct stream_filter filter; |
| 1730 | struct stream_filter *one; |
| 1731 | struct stream_filter *two; |
| 1732 | char buf[FILTER_BUFFER]; |
| 1733 | int end, ptr; |
| 1734 | }; |
| 1735 | |
| 1736 | static int cascade_filter_fn(struct stream_filter *filter, |
| 1737 | const char *input, size_t *isize_p, |
| 1738 | char *output, size_t *osize_p) |
| 1739 | { |
| 1740 | struct cascade_filter *cas = (struct cascade_filter *) filter; |
| 1741 | size_t filled = 0; |
| 1742 | size_t sz = *osize_p; |
| 1743 | size_t to_feed, remaining; |
| 1744 | |
| 1745 | /* |
| 1746 | * input -- (one) --> buf -- (two) --> output |
| 1747 | */ |
| 1748 | while (filled < sz) { |
| 1749 | remaining = sz - filled; |
| 1750 | |
| 1751 | /* do we already have something to feed two with? */ |
| 1752 | if (cas->ptr < cas->end) { |
| 1753 | to_feed = cas->end - cas->ptr; |
| 1754 | if (stream_filter(cas->two, |
| 1755 | cas->buf + cas->ptr, &to_feed, |
| 1756 | output + filled, &remaining)) |
| 1757 | return -1; |
| 1758 | cas->ptr += (cas->end - cas->ptr) - to_feed; |
| 1759 | filled = sz - remaining; |
| 1760 | continue; |
| 1761 | } |
| 1762 | |
| 1763 | /* feed one from upstream and have it emit into our buffer */ |
| 1764 | to_feed = input ? *isize_p : 0; |
| 1765 | if (input && !to_feed) |
| 1766 | break; |
| 1767 | remaining = sizeof(cas->buf); |
| 1768 | if (stream_filter(cas->one, |
| 1769 | input, &to_feed, |
| 1770 | cas->buf, &remaining)) |
| 1771 | return -1; |
| 1772 | cas->end = sizeof(cas->buf) - remaining; |
| 1773 | cas->ptr = 0; |
| 1774 | if (input) { |
| 1775 | size_t fed = *isize_p - to_feed; |
| 1776 | *isize_p -= fed; |
| 1777 | input += fed; |
| 1778 | } |
| 1779 | |
| 1780 | /* do we know that we drained one completely? */ |
| 1781 | if (input || cas->end) |
| 1782 | continue; |
| 1783 | |
| 1784 | /* tell two to drain; we have nothing more to give it */ |
| 1785 | to_feed = 0; |
| 1786 | remaining = sz - filled; |
| 1787 | if (stream_filter(cas->two, |
| 1788 | NULL, &to_feed, |
| 1789 | output + filled, &remaining)) |
| 1790 | return -1; |
| 1791 | if (remaining == (sz - filled)) |
| 1792 | break; /* completely drained two */ |
| 1793 | filled = sz - remaining; |
| 1794 | } |
| 1795 | *osize_p -= filled; |
| 1796 | return 0; |
| 1797 | } |
| 1798 | |
| 1799 | static void cascade_free_fn(struct stream_filter *filter) |
| 1800 | { |
| 1801 | struct cascade_filter *cas = (struct cascade_filter *)filter; |
| 1802 | free_stream_filter(cas->one); |
| 1803 | free_stream_filter(cas->two); |
| 1804 | free(filter); |
| 1805 | } |
| 1806 | |
| 1807 | static struct stream_filter_vtbl cascade_vtbl = { |
| 1808 | .filter = cascade_filter_fn, |
| 1809 | .free = cascade_free_fn, |
| 1810 | }; |
| 1811 | |
| 1812 | static struct stream_filter *cascade_filter(struct stream_filter *one, |
| 1813 | struct stream_filter *two) |
| 1814 | { |
| 1815 | struct cascade_filter *cascade; |
| 1816 | |
| 1817 | if (!one || is_null_stream_filter(one)) |
| 1818 | return two; |
| 1819 | if (!two || is_null_stream_filter(two)) |
| 1820 | return one; |
| 1821 | |
| 1822 | cascade = xmalloc(sizeof(*cascade)); |
| 1823 | cascade->one = one; |
| 1824 | cascade->two = two; |
| 1825 | cascade->end = cascade->ptr = 0; |
| 1826 | cascade->filter.vtbl = &cascade_vtbl; |
| 1827 | return (struct stream_filter *)cascade; |
| 1828 | } |
| 1829 | |
| 1830 | /* |
| 1831 | * ident filter |
| 1832 | */ |
| 1833 | #define IDENT_DRAINING (-1) |
| 1834 | #define IDENT_SKIPPING (-2) |
| 1835 | struct ident_filter { |
| 1836 | struct stream_filter filter; |
| 1837 | struct strbuf left; |
| 1838 | int state; |
| 1839 | char ident[GIT_MAX_HEXSZ + 5]; /* ": x40 $" */ |
| 1840 | }; |
| 1841 | |
| 1842 | static int is_foreign_ident(const char *str) |
| 1843 | { |
| 1844 | int i; |
| 1845 | |
| 1846 | if (!skip_prefix(str, "$Id: ", &str)) |
| 1847 | return 0; |
| 1848 | for (i = 0; str[i]; i++) { |
| 1849 | if (isspace(str[i]) && str[i+1] != '$') |
| 1850 | return 1; |
| 1851 | } |
| 1852 | return 0; |
| 1853 | } |
| 1854 | |
| 1855 | static void ident_drain(struct ident_filter *ident, char **output_p, size_t *osize_p) |
| 1856 | { |
| 1857 | size_t to_drain = ident->left.len; |
| 1858 | |
| 1859 | if (*osize_p < to_drain) |
| 1860 | to_drain = *osize_p; |
| 1861 | if (to_drain) { |
| 1862 | memcpy(*output_p, ident->left.buf, to_drain); |
| 1863 | strbuf_remove(&ident->left, 0, to_drain); |
| 1864 | *output_p += to_drain; |
| 1865 | *osize_p -= to_drain; |
| 1866 | } |
| 1867 | if (!ident->left.len) |
| 1868 | ident->state = 0; |
| 1869 | } |
| 1870 | |
| 1871 | static int ident_filter_fn(struct stream_filter *filter, |
| 1872 | const char *input, size_t *isize_p, |
| 1873 | char *output, size_t *osize_p) |
| 1874 | { |
| 1875 | struct ident_filter *ident = (struct ident_filter *)filter; |
| 1876 | static const char head[] = "$Id"; |
| 1877 | |
| 1878 | if (!input) { |
| 1879 | /* drain upon eof */ |
| 1880 | switch (ident->state) { |
| 1881 | default: |
| 1882 | strbuf_add(&ident->left, head, ident->state); |
| 1883 | /* fallthrough */ |
| 1884 | case IDENT_SKIPPING: |
| 1885 | /* fallthrough */ |
| 1886 | case IDENT_DRAINING: |
| 1887 | ident_drain(ident, &output, osize_p); |
| 1888 | } |
| 1889 | return 0; |
| 1890 | } |
| 1891 | |
| 1892 | while (*isize_p || (ident->state == IDENT_DRAINING)) { |
| 1893 | int ch; |
| 1894 | |
| 1895 | if (ident->state == IDENT_DRAINING) { |
| 1896 | ident_drain(ident, &output, osize_p); |
| 1897 | if (!*osize_p) |
| 1898 | break; |
| 1899 | continue; |
| 1900 | } |
| 1901 | |
| 1902 | ch = *(input++); |
| 1903 | (*isize_p)--; |
| 1904 | |
| 1905 | if (ident->state == IDENT_SKIPPING) { |
| 1906 | /* |
| 1907 | * Skipping until '$' or LF, but keeping them |
| 1908 | * in case it is a foreign ident. |
| 1909 | */ |
| 1910 | strbuf_addch(&ident->left, ch); |
| 1911 | if (ch != '\n' && ch != '$') |
| 1912 | continue; |
| 1913 | if (ch == '$' && !is_foreign_ident(ident->left.buf)) { |
| 1914 | strbuf_setlen(&ident->left, sizeof(head) - 1); |
| 1915 | strbuf_addstr(&ident->left, ident->ident); |
| 1916 | } |
| 1917 | ident->state = IDENT_DRAINING; |
| 1918 | continue; |
| 1919 | } |
| 1920 | |
| 1921 | if (ident->state < sizeof(head) && |
| 1922 | head[ident->state] == ch) { |
| 1923 | ident->state++; |
| 1924 | continue; |
| 1925 | } |
| 1926 | |
| 1927 | if (ident->state) |
| 1928 | strbuf_add(&ident->left, head, ident->state); |
| 1929 | if (ident->state == sizeof(head) - 1) { |
| 1930 | if (ch != ':' && ch != '$') { |
| 1931 | strbuf_addch(&ident->left, ch); |
| 1932 | ident->state = 0; |
| 1933 | continue; |
| 1934 | } |
| 1935 | |
| 1936 | if (ch == ':') { |
| 1937 | strbuf_addch(&ident->left, ch); |
| 1938 | ident->state = IDENT_SKIPPING; |
| 1939 | } else { |
| 1940 | strbuf_addstr(&ident->left, ident->ident); |
| 1941 | ident->state = IDENT_DRAINING; |
| 1942 | } |
| 1943 | continue; |
| 1944 | } |
| 1945 | |
| 1946 | strbuf_addch(&ident->left, ch); |
| 1947 | ident->state = IDENT_DRAINING; |
| 1948 | } |
| 1949 | return 0; |
| 1950 | } |
| 1951 | |
| 1952 | static void ident_free_fn(struct stream_filter *filter) |
| 1953 | { |
| 1954 | struct ident_filter *ident = (struct ident_filter *)filter; |
| 1955 | strbuf_release(&ident->left); |
| 1956 | free(filter); |
| 1957 | } |
| 1958 | |
| 1959 | static struct stream_filter_vtbl ident_vtbl = { |
| 1960 | .filter = ident_filter_fn, |
| 1961 | .free = ident_free_fn, |
| 1962 | }; |
| 1963 | |
| 1964 | static struct stream_filter *ident_filter(const struct object_id *oid) |
| 1965 | { |
| 1966 | struct ident_filter *ident = xmalloc(sizeof(*ident)); |
| 1967 | |
| 1968 | xsnprintf(ident->ident, sizeof(ident->ident), |
| 1969 | ": %s $", oid_to_hex(oid)); |
| 1970 | strbuf_init(&ident->left, 0); |
| 1971 | ident->filter.vtbl = &ident_vtbl; |
| 1972 | ident->state = 0; |
| 1973 | return (struct stream_filter *)ident; |
| 1974 | } |
| 1975 | |
| 1976 | /* |
| 1977 | * Return an appropriately constructed filter for the given ca, or NULL if |
| 1978 | * the contents cannot be filtered without reading the whole thing |
| 1979 | * in-core. |
| 1980 | * |
| 1981 | * Note that you would be crazy to set CRLF, smudge/clean or ident to a |
| 1982 | * large binary blob you would want us not to slurp into the memory! |
| 1983 | */ |
| 1984 | struct stream_filter *get_stream_filter_ca(const struct conv_attrs *ca, |
| 1985 | const struct object_id *oid) |
| 1986 | { |
| 1987 | struct stream_filter *filter = NULL; |
| 1988 | |
| 1989 | if (classify_conv_attrs(ca) != CA_CLASS_STREAMABLE) |
| 1990 | return NULL; |
| 1991 | |
| 1992 | if (ca->ident) |
| 1993 | filter = ident_filter(oid); |
| 1994 | |
| 1995 | if (output_eol(ca->crlf_action) == EOL_CRLF) |
| 1996 | filter = cascade_filter(filter, lf_to_crlf_filter()); |
| 1997 | else |
| 1998 | filter = cascade_filter(filter, &null_filter_singleton); |
| 1999 | |
| 2000 | return filter; |
| 2001 | } |
| 2002 | |
| 2003 | struct stream_filter *get_stream_filter(struct index_state *istate, |
| 2004 | const char *path, |
| 2005 | const struct object_id *oid) |
| 2006 | { |
| 2007 | struct conv_attrs ca; |
| 2008 | convert_attrs(istate, &ca, path); |
| 2009 | return get_stream_filter_ca(&ca, oid); |
| 2010 | } |
| 2011 | |
| 2012 | void free_stream_filter(struct stream_filter *filter) |
| 2013 | { |
| 2014 | filter->vtbl->free(filter); |
| 2015 | } |
| 2016 | |
| 2017 | int stream_filter(struct stream_filter *filter, |
| 2018 | const char *input, size_t *isize_p, |
| 2019 | char *output, size_t *osize_p) |
| 2020 | { |
| 2021 | return filter->vtbl->filter(filter, input, isize_p, output, osize_p); |
| 2022 | } |
| 2023 | |
| 2024 | void init_checkout_metadata(struct checkout_metadata *meta, const char *refname, |
| 2025 | const struct object_id *treeish, |
| 2026 | const struct object_id *blob) |
| 2027 | { |
| 2028 | memset(meta, 0, sizeof(*meta)); |
| 2029 | if (refname) |
| 2030 | meta->refname = refname; |
| 2031 | if (treeish) |
| 2032 | oidcpy(&meta->treeish, treeish); |
| 2033 | if (blob) |
| 2034 | oidcpy(&meta->blob, blob); |
| 2035 | } |
| 2036 | |
| 2037 | void clone_checkout_metadata(struct checkout_metadata *dst, |
| 2038 | const struct checkout_metadata *src, |
| 2039 | const struct object_id *blob) |
| 2040 | { |
| 2041 | memcpy(dst, src, sizeof(*dst)); |
| 2042 | if (blob) |
| 2043 | oidcpy(&dst->blob, blob); |
| 2044 | } |
| 2045 | |
| 2046 | enum conv_attrs_classification classify_conv_attrs(const struct conv_attrs *ca) |
| 2047 | { |
| 2048 | if (ca->drv) { |
| 2049 | if (ca->drv->process) |
| 2050 | return CA_CLASS_INCORE_PROCESS; |
| 2051 | if (ca->drv->smudge || ca->drv->clean) |
| 2052 | return CA_CLASS_INCORE_FILTER; |
| 2053 | } |
| 2054 | |
| 2055 | if (ca->working_tree_encoding) |
| 2056 | return CA_CLASS_INCORE; |
| 2057 | |
| 2058 | if (ca->crlf_action == CRLF_AUTO || ca->crlf_action == CRLF_AUTO_CRLF) |
| 2059 | return CA_CLASS_INCORE; |
| 2060 | |
| 2061 | return CA_CLASS_STREAMABLE; |
| 2062 | } |