| 1 | /* |
| 2 | * LibXDiff by Davide Libenzi ( File Differential Library ) |
| 3 | * Copyright (C) 2003 Davide Libenzi |
| 4 | * |
| 5 | * This library is free software; you can redistribute it and/or |
| 6 | * modify it under the terms of the GNU Lesser General Public |
| 7 | * License as published by the Free Software Foundation; either |
| 8 | * version 2.1 of the License, or (at your option) any later version. |
| 9 | * |
| 10 | * This library is distributed in the hope that it will be useful, |
| 11 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 12 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 13 | * Lesser General Public License for more details. |
| 14 | * |
| 15 | * You should have received a copy of the GNU Lesser General Public |
| 16 | * License along with this library; if not, see |
| 17 | * <http://www.gnu.org/licenses/>. |
| 18 | * |
| 19 | * Davide Libenzi <davidel@xmailserver.org> |
| 20 | * |
| 21 | */ |
| 22 | |
| 23 | #include "xinclude.h" |
| 24 | |
| 25 | static size_t get_hash(xdfile_t *xdf, long index) |
| 26 | { |
| 27 | return xdf->recs[xdf->reference_index[index]].minimal_perfect_hash; |
| 28 | } |
| 29 | |
| 30 | #define XDL_MAX_COST_MIN 256 |
| 31 | #define XDL_HEUR_MIN_COST 256 |
| 32 | #define XDL_LINE_MAX (long)((1UL << (CHAR_BIT * sizeof(long) - 1)) - 1) |
| 33 | #define XDL_SNAKE_CNT 20 |
| 34 | #define XDL_K_HEUR 4 |
| 35 | |
| 36 | typedef struct s_xdpsplit { |
| 37 | long i1, i2; |
| 38 | int min_lo, min_hi; |
| 39 | } xdpsplit_t; |
| 40 | |
| 41 | /* |
| 42 | * See "An O(ND) Difference Algorithm and its Variations", by Eugene Myers. |
| 43 | * Basically considers a "box" (off1, off2, lim1, lim2) and scan from both |
| 44 | * the forward diagonal starting from (off1, off2) and the backward diagonal |
| 45 | * starting from (lim1, lim2). If the K values on the same diagonal crosses |
| 46 | * returns the furthest point of reach. We might encounter expensive edge cases |
| 47 | * using this algorithm, so a little bit of heuristic is needed to cut the |
| 48 | * search and to return a suboptimal point. |
| 49 | */ |
| 50 | static long xdl_split(xdfile_t *xdf1, long off1, long lim1, |
| 51 | xdfile_t *xdf2, long off2, long lim2, |
| 52 | long *kvdf, long *kvdb, int need_min, xdpsplit_t *spl, |
| 53 | xdalgoenv_t *xenv) { |
| 54 | long dmin = off1 - lim2, dmax = lim1 - off2; |
| 55 | long fmid = off1 - off2, bmid = lim1 - lim2; |
| 56 | long odd = (fmid - bmid) & 1; |
| 57 | long fmin = fmid, fmax = fmid; |
| 58 | long bmin = bmid, bmax = bmid; |
| 59 | long ec, d, i1, i2, prev1, best, dd, v, k; |
| 60 | |
| 61 | /* |
| 62 | * Set initial diagonal values for both forward and backward path. |
| 63 | */ |
| 64 | kvdf[fmid] = off1; |
| 65 | kvdb[bmid] = lim1; |
| 66 | |
| 67 | for (ec = 1;; ec++) { |
| 68 | int got_snake = 0; |
| 69 | |
| 70 | /* |
| 71 | * We need to extend the diagonal "domain" by one. If the next |
| 72 | * values exits the box boundaries we need to change it in the |
| 73 | * opposite direction because (max - min) must be a power of |
| 74 | * two. |
| 75 | * |
| 76 | * Also we initialize the external K value to -1 so that we can |
| 77 | * avoid extra conditions in the check inside the core loop. |
| 78 | */ |
| 79 | if (fmin > dmin) |
| 80 | kvdf[--fmin - 1] = -1; |
| 81 | else |
| 82 | ++fmin; |
| 83 | if (fmax < dmax) |
| 84 | kvdf[++fmax + 1] = -1; |
| 85 | else |
| 86 | --fmax; |
| 87 | |
| 88 | for (d = fmax; d >= fmin; d -= 2) { |
| 89 | if (kvdf[d - 1] >= kvdf[d + 1]) |
| 90 | i1 = kvdf[d - 1] + 1; |
| 91 | else |
| 92 | i1 = kvdf[d + 1]; |
| 93 | prev1 = i1; |
| 94 | i2 = i1 - d; |
| 95 | for (; i1 < lim1 && i2 < lim2 && get_hash(xdf1, i1) == get_hash(xdf2, i2); i1++, i2++); |
| 96 | if (i1 - prev1 > xenv->snake_cnt) |
| 97 | got_snake = 1; |
| 98 | kvdf[d] = i1; |
| 99 | if (odd && bmin <= d && d <= bmax && kvdb[d] <= i1) { |
| 100 | spl->i1 = i1; |
| 101 | spl->i2 = i2; |
| 102 | spl->min_lo = spl->min_hi = 1; |
| 103 | return ec; |
| 104 | } |
| 105 | } |
| 106 | |
| 107 | /* |
| 108 | * We need to extend the diagonal "domain" by one. If the next |
| 109 | * values exits the box boundaries we need to change it in the |
| 110 | * opposite direction because (max - min) must be a power of |
| 111 | * two. |
| 112 | * |
| 113 | * Also we initialize the external K value to -1 so that we can |
| 114 | * avoid extra conditions in the check inside the core loop. |
| 115 | */ |
| 116 | if (bmin > dmin) |
| 117 | kvdb[--bmin - 1] = XDL_LINE_MAX; |
| 118 | else |
| 119 | ++bmin; |
| 120 | if (bmax < dmax) |
| 121 | kvdb[++bmax + 1] = XDL_LINE_MAX; |
| 122 | else |
| 123 | --bmax; |
| 124 | |
| 125 | for (d = bmax; d >= bmin; d -= 2) { |
| 126 | if (kvdb[d - 1] < kvdb[d + 1]) |
| 127 | i1 = kvdb[d - 1]; |
| 128 | else |
| 129 | i1 = kvdb[d + 1] - 1; |
| 130 | prev1 = i1; |
| 131 | i2 = i1 - d; |
| 132 | for (; i1 > off1 && i2 > off2 && get_hash(xdf1, i1 - 1) == get_hash(xdf2, i2 - 1); i1--, i2--); |
| 133 | if (prev1 - i1 > xenv->snake_cnt) |
| 134 | got_snake = 1; |
| 135 | kvdb[d] = i1; |
| 136 | if (!odd && fmin <= d && d <= fmax && i1 <= kvdf[d]) { |
| 137 | spl->i1 = i1; |
| 138 | spl->i2 = i2; |
| 139 | spl->min_lo = spl->min_hi = 1; |
| 140 | return ec; |
| 141 | } |
| 142 | } |
| 143 | |
| 144 | if (need_min) |
| 145 | continue; |
| 146 | |
| 147 | /* |
| 148 | * If the edit cost is above the heuristic trigger and if |
| 149 | * we got a good snake, we sample current diagonals to see |
| 150 | * if some of them have reached an "interesting" path. Our |
| 151 | * measure is a function of the distance from the diagonal |
| 152 | * corner (i1 + i2) penalized with the distance from the |
| 153 | * mid diagonal itself. If this value is above the current |
| 154 | * edit cost times a magic factor (XDL_K_HEUR) we consider |
| 155 | * it interesting. |
| 156 | */ |
| 157 | if (got_snake && ec > xenv->heur_min) { |
| 158 | for (best = 0, d = fmax; d >= fmin; d -= 2) { |
| 159 | dd = d > fmid ? d - fmid: fmid - d; |
| 160 | i1 = kvdf[d]; |
| 161 | i2 = i1 - d; |
| 162 | v = (i1 - off1) + (i2 - off2) - dd; |
| 163 | |
| 164 | if (v > XDL_K_HEUR * ec && v > best && |
| 165 | off1 + xenv->snake_cnt <= i1 && i1 < lim1 && |
| 166 | off2 + xenv->snake_cnt <= i2 && i2 < lim2) { |
| 167 | for (k = 1; get_hash(xdf1, i1 - k) == get_hash(xdf2, i2 - k); k++) |
| 168 | if (k == xenv->snake_cnt) { |
| 169 | best = v; |
| 170 | spl->i1 = i1; |
| 171 | spl->i2 = i2; |
| 172 | break; |
| 173 | } |
| 174 | } |
| 175 | } |
| 176 | if (best > 0) { |
| 177 | spl->min_lo = 1; |
| 178 | spl->min_hi = 0; |
| 179 | return ec; |
| 180 | } |
| 181 | |
| 182 | for (best = 0, d = bmax; d >= bmin; d -= 2) { |
| 183 | dd = d > bmid ? d - bmid: bmid - d; |
| 184 | i1 = kvdb[d]; |
| 185 | i2 = i1 - d; |
| 186 | v = (lim1 - i1) + (lim2 - i2) - dd; |
| 187 | |
| 188 | if (v > XDL_K_HEUR * ec && v > best && |
| 189 | off1 < i1 && i1 <= lim1 - xenv->snake_cnt && |
| 190 | off2 < i2 && i2 <= lim2 - xenv->snake_cnt) { |
| 191 | for (k = 0; get_hash(xdf1, i1 + k) == get_hash(xdf2, i2 + k); k++) |
| 192 | if (k == xenv->snake_cnt - 1) { |
| 193 | best = v; |
| 194 | spl->i1 = i1; |
| 195 | spl->i2 = i2; |
| 196 | break; |
| 197 | } |
| 198 | } |
| 199 | } |
| 200 | if (best > 0) { |
| 201 | spl->min_lo = 0; |
| 202 | spl->min_hi = 1; |
| 203 | return ec; |
| 204 | } |
| 205 | } |
| 206 | |
| 207 | /* |
| 208 | * Enough is enough. We spent too much time here and now we |
| 209 | * collect the furthest reaching path using the (i1 + i2) |
| 210 | * measure. |
| 211 | */ |
| 212 | if (ec >= xenv->mxcost) { |
| 213 | long fbest, fbest1, bbest, bbest1; |
| 214 | |
| 215 | fbest = fbest1 = -1; |
| 216 | for (d = fmax; d >= fmin; d -= 2) { |
| 217 | i1 = XDL_MIN(kvdf[d], lim1); |
| 218 | i2 = i1 - d; |
| 219 | if (lim2 < i2) { |
| 220 | i1 = lim2 + d; |
| 221 | i2 = lim2; |
| 222 | } |
| 223 | if (fbest < i1 + i2) { |
| 224 | fbest = i1 + i2; |
| 225 | fbest1 = i1; |
| 226 | } |
| 227 | } |
| 228 | |
| 229 | bbest = bbest1 = XDL_LINE_MAX; |
| 230 | for (d = bmax; d >= bmin; d -= 2) { |
| 231 | i1 = XDL_MAX(off1, kvdb[d]); |
| 232 | i2 = i1 - d; |
| 233 | if (i2 < off2) { |
| 234 | i1 = off2 + d; |
| 235 | i2 = off2; |
| 236 | } |
| 237 | if (i1 + i2 < bbest) { |
| 238 | bbest = i1 + i2; |
| 239 | bbest1 = i1; |
| 240 | } |
| 241 | } |
| 242 | |
| 243 | if ((lim1 + lim2) - bbest < fbest - (off1 + off2)) { |
| 244 | spl->i1 = fbest1; |
| 245 | spl->i2 = fbest - fbest1; |
| 246 | spl->min_lo = 1; |
| 247 | spl->min_hi = 0; |
| 248 | } else { |
| 249 | spl->i1 = bbest1; |
| 250 | spl->i2 = bbest - bbest1; |
| 251 | spl->min_lo = 0; |
| 252 | spl->min_hi = 1; |
| 253 | } |
| 254 | return ec; |
| 255 | } |
| 256 | } |
| 257 | } |
| 258 | |
| 259 | |
| 260 | /* |
| 261 | * Rule: "Divide et Impera" (divide & conquer). Recursively split the box in |
| 262 | * sub-boxes by calling the box splitting function. Note that the real job |
| 263 | * (marking changed lines) is done in the two boundary reaching checks. |
| 264 | */ |
| 265 | int xdl_recs_cmp(xdfile_t *xdf1, long off1, long lim1, |
| 266 | xdfile_t *xdf2, long off2, long lim2, |
| 267 | long *kvdf, long *kvdb, int need_min, xdalgoenv_t *xenv) { |
| 268 | |
| 269 | /* |
| 270 | * Shrink the box by walking through each diagonal snake (SW and NE). |
| 271 | */ |
| 272 | for (; off1 < lim1 && off2 < lim2 && get_hash(xdf1, off1) == get_hash(xdf2, off2); off1++, off2++); |
| 273 | for (; off1 < lim1 && off2 < lim2 && get_hash(xdf1, lim1 - 1) == get_hash(xdf2, lim2 - 1); lim1--, lim2--); |
| 274 | |
| 275 | /* |
| 276 | * If one dimension is empty, then all records on the other one must |
| 277 | * be obviously changed. |
| 278 | */ |
| 279 | if (off1 == lim1) { |
| 280 | for (; off2 < lim2; off2++) |
| 281 | xdf2->changed[xdf2->reference_index[off2]] = true; |
| 282 | } else if (off2 == lim2) { |
| 283 | for (; off1 < lim1; off1++) |
| 284 | xdf1->changed[xdf1->reference_index[off1]] = true; |
| 285 | } else { |
| 286 | xdpsplit_t spl; |
| 287 | spl.i1 = spl.i2 = 0; |
| 288 | |
| 289 | /* |
| 290 | * Divide ... |
| 291 | */ |
| 292 | if (xdl_split(xdf1, off1, lim1, xdf2, off2, lim2, kvdf, kvdb, |
| 293 | need_min, &spl, xenv) < 0) { |
| 294 | |
| 295 | return -1; |
| 296 | } |
| 297 | |
| 298 | /* |
| 299 | * ... et Impera. |
| 300 | */ |
| 301 | if (xdl_recs_cmp(xdf1, off1, spl.i1, xdf2, off2, spl.i2, |
| 302 | kvdf, kvdb, spl.min_lo, xenv) < 0 || |
| 303 | xdl_recs_cmp(xdf1, spl.i1, lim1, xdf2, spl.i2, lim2, |
| 304 | kvdf, kvdb, spl.min_hi, xenv) < 0) { |
| 305 | |
| 306 | return -1; |
| 307 | } |
| 308 | } |
| 309 | |
| 310 | return 0; |
| 311 | } |
| 312 | |
| 313 | |
| 314 | int xdl_do_diff(mmfile_t *mf1, mmfile_t *mf2, xpparam_t const *xpp, |
| 315 | xdfenv_t *xe) { |
| 316 | long ndiags; |
| 317 | long *kvd, *kvdf, *kvdb; |
| 318 | xdalgoenv_t xenv; |
| 319 | int res; |
| 320 | |
| 321 | if (xdl_prepare_env(mf1, mf2, xpp, xe) < 0) |
| 322 | return -1; |
| 323 | |
| 324 | if (XDF_DIFF_ALG(xpp->flags) == XDF_PATIENCE_DIFF) { |
| 325 | res = xdl_do_patience_diff(xpp, xe); |
| 326 | goto out; |
| 327 | } |
| 328 | |
| 329 | if (XDF_DIFF_ALG(xpp->flags) == XDF_HISTOGRAM_DIFF) { |
| 330 | res = xdl_do_histogram_diff(xpp, xe); |
| 331 | goto out; |
| 332 | } |
| 333 | |
| 334 | /* |
| 335 | * Allocate and setup K vectors to be used by the differential |
| 336 | * algorithm. |
| 337 | * |
| 338 | * One is to store the forward path and one to store the backward path. |
| 339 | */ |
| 340 | ndiags = xe->xdf1.nreff + xe->xdf2.nreff + 3; |
| 341 | if (!XDL_ALLOC_ARRAY(kvd, 2 * ndiags + 2)) { |
| 342 | |
| 343 | xdl_free_env(xe); |
| 344 | return -1; |
| 345 | } |
| 346 | kvdf = kvd; |
| 347 | kvdb = kvdf + ndiags; |
| 348 | kvdf += xe->xdf2.nreff + 1; |
| 349 | kvdb += xe->xdf2.nreff + 1; |
| 350 | |
| 351 | xenv.mxcost = (long)xdl_bogosqrt((uint64_t)ndiags); |
| 352 | if (xenv.mxcost < XDL_MAX_COST_MIN) |
| 353 | xenv.mxcost = XDL_MAX_COST_MIN; |
| 354 | xenv.snake_cnt = XDL_SNAKE_CNT; |
| 355 | xenv.heur_min = XDL_HEUR_MIN_COST; |
| 356 | |
| 357 | res = xdl_recs_cmp(&xe->xdf1, 0, xe->xdf1.nreff, &xe->xdf2, 0, xe->xdf2.nreff, |
| 358 | kvdf, kvdb, (xpp->flags & XDF_NEED_MINIMAL) != 0, |
| 359 | &xenv); |
| 360 | xdl_free(kvd); |
| 361 | out: |
| 362 | if (res < 0) |
| 363 | xdl_free_env(xe); |
| 364 | |
| 365 | return res; |
| 366 | } |
| 367 | |
| 368 | |
| 369 | static xdchange_t *xdl_add_change(xdchange_t *xscr, long i1, long i2, long chg1, long chg2) { |
| 370 | xdchange_t *xch; |
| 371 | |
| 372 | if (!(xch = (xdchange_t *) xdl_malloc(sizeof(xdchange_t)))) |
| 373 | return NULL; |
| 374 | |
| 375 | xch->next = xscr; |
| 376 | xch->i1 = i1; |
| 377 | xch->i2 = i2; |
| 378 | xch->chg1 = chg1; |
| 379 | xch->chg2 = chg2; |
| 380 | xch->ignore = 0; |
| 381 | |
| 382 | return xch; |
| 383 | } |
| 384 | |
| 385 | |
| 386 | static int recs_match(xrecord_t *rec1, xrecord_t *rec2) |
| 387 | { |
| 388 | return rec1->minimal_perfect_hash == rec2->minimal_perfect_hash; |
| 389 | } |
| 390 | |
| 391 | /* |
| 392 | * If a line is indented more than this, get_indent() just returns this value. |
| 393 | * This avoids having to do absurd amounts of work for data that are not |
| 394 | * human-readable text, and also ensures that the output of get_indent fits |
| 395 | * within an int. |
| 396 | */ |
| 397 | #define MAX_INDENT 200 |
| 398 | |
| 399 | /* |
| 400 | * Return the amount of indentation of the specified line, treating TAB as 8 |
| 401 | * columns. Return -1 if line is empty or contains only whitespace. Clamp the |
| 402 | * output value at MAX_INDENT. |
| 403 | */ |
| 404 | static int get_indent(xrecord_t *rec) |
| 405 | { |
| 406 | int ret = 0; |
| 407 | |
| 408 | for (size_t i = 0; i < rec->size; i++) { |
| 409 | char c = (char) rec->ptr[i]; |
| 410 | |
| 411 | if (!XDL_ISSPACE(c)) |
| 412 | return ret; |
| 413 | else if (c == ' ') |
| 414 | ret += 1; |
| 415 | else if (c == '\t') |
| 416 | ret += 8 - ret % 8; |
| 417 | /* ignore other whitespace characters */ |
| 418 | |
| 419 | if (ret >= MAX_INDENT) |
| 420 | return MAX_INDENT; |
| 421 | } |
| 422 | |
| 423 | /* The line contains only whitespace. */ |
| 424 | return -1; |
| 425 | } |
| 426 | |
| 427 | /* |
| 428 | * If more than this number of consecutive blank rows are found, just return |
| 429 | * this value. This avoids requiring O(N^2) work for pathological cases, and |
| 430 | * also ensures that the output of score_split fits in an int. |
| 431 | */ |
| 432 | #define MAX_BLANKS 20 |
| 433 | |
| 434 | /* Characteristics measured about a hypothetical split position. */ |
| 435 | struct split_measurement { |
| 436 | /* |
| 437 | * Is the split at the end of the file (aside from any blank lines)? |
| 438 | */ |
| 439 | int end_of_file; |
| 440 | |
| 441 | /* |
| 442 | * How much is the line immediately following the split indented (or -1 |
| 443 | * if the line is blank): |
| 444 | */ |
| 445 | int indent; |
| 446 | |
| 447 | /* |
| 448 | * How many consecutive lines above the split are blank? |
| 449 | */ |
| 450 | int pre_blank; |
| 451 | |
| 452 | /* |
| 453 | * How much is the nearest non-blank line above the split indented (or |
| 454 | * -1 if there is no such line)? |
| 455 | */ |
| 456 | int pre_indent; |
| 457 | |
| 458 | /* |
| 459 | * How many lines after the line following the split are blank? |
| 460 | */ |
| 461 | int post_blank; |
| 462 | |
| 463 | /* |
| 464 | * How much is the nearest non-blank line after the line following the |
| 465 | * split indented (or -1 if there is no such line)? |
| 466 | */ |
| 467 | int post_indent; |
| 468 | }; |
| 469 | |
| 470 | struct split_score { |
| 471 | /* The effective indent of this split (smaller is preferred). */ |
| 472 | int effective_indent; |
| 473 | |
| 474 | /* Penalty for this split (smaller is preferred). */ |
| 475 | int penalty; |
| 476 | }; |
| 477 | |
| 478 | /* |
| 479 | * Fill m with information about a hypothetical split of xdf above line split. |
| 480 | */ |
| 481 | static void measure_split(const xdfile_t *xdf, long split, |
| 482 | struct split_measurement *m) |
| 483 | { |
| 484 | long i; |
| 485 | |
| 486 | if (split >= (long)xdf->nrec) { |
| 487 | m->end_of_file = 1; |
| 488 | m->indent = -1; |
| 489 | } else { |
| 490 | m->end_of_file = 0; |
| 491 | m->indent = get_indent(&xdf->recs[split]); |
| 492 | } |
| 493 | |
| 494 | m->pre_blank = 0; |
| 495 | m->pre_indent = -1; |
| 496 | for (i = split - 1; i >= 0; i--) { |
| 497 | m->pre_indent = get_indent(&xdf->recs[i]); |
| 498 | if (m->pre_indent != -1) |
| 499 | break; |
| 500 | m->pre_blank += 1; |
| 501 | if (m->pre_blank == MAX_BLANKS) { |
| 502 | m->pre_indent = 0; |
| 503 | break; |
| 504 | } |
| 505 | } |
| 506 | |
| 507 | m->post_blank = 0; |
| 508 | m->post_indent = -1; |
| 509 | for (i = split + 1; i < (long)xdf->nrec; i++) { |
| 510 | m->post_indent = get_indent(&xdf->recs[i]); |
| 511 | if (m->post_indent != -1) |
| 512 | break; |
| 513 | m->post_blank += 1; |
| 514 | if (m->post_blank == MAX_BLANKS) { |
| 515 | m->post_indent = 0; |
| 516 | break; |
| 517 | } |
| 518 | } |
| 519 | } |
| 520 | |
| 521 | /* |
| 522 | * The empirically-determined weight factors used by score_split() below. |
| 523 | * Larger values means that the position is a less favorable place to split. |
| 524 | * |
| 525 | * Note that scores are only ever compared against each other, so multiplying |
| 526 | * all of these weight/penalty values by the same factor wouldn't change the |
| 527 | * heuristic's behavior. Still, we need to set that arbitrary scale *somehow*. |
| 528 | * In practice, these numbers are chosen to be large enough that they can be |
| 529 | * adjusted relative to each other with sufficient precision despite using |
| 530 | * integer math. |
| 531 | */ |
| 532 | |
| 533 | /* Penalty if there are no non-blank lines before the split */ |
| 534 | #define START_OF_FILE_PENALTY 1 |
| 535 | |
| 536 | /* Penalty if there are no non-blank lines after the split */ |
| 537 | #define END_OF_FILE_PENALTY 21 |
| 538 | |
| 539 | /* Multiplier for the number of blank lines around the split */ |
| 540 | #define TOTAL_BLANK_WEIGHT (-30) |
| 541 | |
| 542 | /* Multiplier for the number of blank lines after the split */ |
| 543 | #define POST_BLANK_WEIGHT 6 |
| 544 | |
| 545 | /* |
| 546 | * Penalties applied if the line is indented more than its predecessor |
| 547 | */ |
| 548 | #define RELATIVE_INDENT_PENALTY (-4) |
| 549 | #define RELATIVE_INDENT_WITH_BLANK_PENALTY 10 |
| 550 | |
| 551 | /* |
| 552 | * Penalties applied if the line is indented less than both its predecessor and |
| 553 | * its successor |
| 554 | */ |
| 555 | #define RELATIVE_OUTDENT_PENALTY 24 |
| 556 | #define RELATIVE_OUTDENT_WITH_BLANK_PENALTY 17 |
| 557 | |
| 558 | /* |
| 559 | * Penalties applied if the line is indented less than its predecessor but not |
| 560 | * less than its successor |
| 561 | */ |
| 562 | #define RELATIVE_DEDENT_PENALTY 23 |
| 563 | #define RELATIVE_DEDENT_WITH_BLANK_PENALTY 17 |
| 564 | |
| 565 | /* |
| 566 | * We only consider whether the sum of the effective indents for splits are |
| 567 | * less than (-1), equal to (0), or greater than (+1) each other. The resulting |
| 568 | * value is multiplied by the following weight and combined with the penalty to |
| 569 | * determine the better of two scores. |
| 570 | */ |
| 571 | #define INDENT_WEIGHT 60 |
| 572 | |
| 573 | /* |
| 574 | * How far do we slide a hunk at most? |
| 575 | */ |
| 576 | #define INDENT_HEURISTIC_MAX_SLIDING 100 |
| 577 | |
| 578 | /* |
| 579 | * Compute a badness score for the hypothetical split whose measurements are |
| 580 | * stored in m. The weight factors were determined empirically using the tools |
| 581 | * and corpus described in |
| 582 | * |
| 583 | * https://github.com/mhagger/diff-slider-tools |
| 584 | * |
| 585 | * Also see that project if you want to improve the weights based on, for |
| 586 | * example, a larger or more diverse corpus. |
| 587 | */ |
| 588 | static void score_add_split(const struct split_measurement *m, struct split_score *s) |
| 589 | { |
| 590 | /* |
| 591 | * A place to accumulate penalty factors (positive makes this index more |
| 592 | * favored): |
| 593 | */ |
| 594 | int post_blank, total_blank, indent, any_blanks; |
| 595 | |
| 596 | if (m->pre_indent == -1 && m->pre_blank == 0) |
| 597 | s->penalty += START_OF_FILE_PENALTY; |
| 598 | |
| 599 | if (m->end_of_file) |
| 600 | s->penalty += END_OF_FILE_PENALTY; |
| 601 | |
| 602 | /* |
| 603 | * Set post_blank to the number of blank lines following the split, |
| 604 | * including the line immediately after the split: |
| 605 | */ |
| 606 | post_blank = (m->indent == -1) ? 1 + m->post_blank : 0; |
| 607 | total_blank = m->pre_blank + post_blank; |
| 608 | |
| 609 | /* Penalties based on nearby blank lines: */ |
| 610 | s->penalty += TOTAL_BLANK_WEIGHT * total_blank; |
| 611 | s->penalty += POST_BLANK_WEIGHT * post_blank; |
| 612 | |
| 613 | if (m->indent != -1) |
| 614 | indent = m->indent; |
| 615 | else |
| 616 | indent = m->post_indent; |
| 617 | |
| 618 | any_blanks = (total_blank != 0); |
| 619 | |
| 620 | /* Note that the effective indent is -1 at the end of the file: */ |
| 621 | s->effective_indent += indent; |
| 622 | |
| 623 | if (indent == -1) { |
| 624 | /* No additional adjustments needed. */ |
| 625 | } else if (m->pre_indent == -1) { |
| 626 | /* No additional adjustments needed. */ |
| 627 | } else if (indent > m->pre_indent) { |
| 628 | /* |
| 629 | * The line is indented more than its predecessor. |
| 630 | */ |
| 631 | s->penalty += any_blanks ? |
| 632 | RELATIVE_INDENT_WITH_BLANK_PENALTY : |
| 633 | RELATIVE_INDENT_PENALTY; |
| 634 | } else if (indent == m->pre_indent) { |
| 635 | /* |
| 636 | * The line has the same indentation level as its predecessor. |
| 637 | * No additional adjustments needed. |
| 638 | */ |
| 639 | } else { |
| 640 | /* |
| 641 | * The line is indented less than its predecessor. It could be |
| 642 | * the block terminator of the previous block, but it could |
| 643 | * also be the start of a new block (e.g., an "else" block, or |
| 644 | * maybe the previous block didn't have a block terminator). |
| 645 | * Try to distinguish those cases based on what comes next: |
| 646 | */ |
| 647 | if (m->post_indent != -1 && m->post_indent > indent) { |
| 648 | /* |
| 649 | * The following line is indented more. So it is likely |
| 650 | * that this line is the start of a block. |
| 651 | */ |
| 652 | s->penalty += any_blanks ? |
| 653 | RELATIVE_OUTDENT_WITH_BLANK_PENALTY : |
| 654 | RELATIVE_OUTDENT_PENALTY; |
| 655 | } else { |
| 656 | /* |
| 657 | * That was probably the end of a block. |
| 658 | */ |
| 659 | s->penalty += any_blanks ? |
| 660 | RELATIVE_DEDENT_WITH_BLANK_PENALTY : |
| 661 | RELATIVE_DEDENT_PENALTY; |
| 662 | } |
| 663 | } |
| 664 | } |
| 665 | |
| 666 | static int score_cmp(struct split_score *s1, struct split_score *s2) |
| 667 | { |
| 668 | /* -1 if s1.effective_indent < s2->effective_indent, etc. */ |
| 669 | int cmp_indents = ((s1->effective_indent > s2->effective_indent) - |
| 670 | (s1->effective_indent < s2->effective_indent)); |
| 671 | |
| 672 | return INDENT_WEIGHT * cmp_indents + (s1->penalty - s2->penalty); |
| 673 | } |
| 674 | |
| 675 | /* |
| 676 | * Represent a group of changed lines in an xdfile_t (i.e., a contiguous group |
| 677 | * of lines that was inserted or deleted from the corresponding version of the |
| 678 | * file). We consider there to be such a group at the beginning of the file, at |
| 679 | * the end of the file, and between any two unchanged lines, though most such |
| 680 | * groups will usually be empty. |
| 681 | * |
| 682 | * If the first line in a group is equal to the line following the group, then |
| 683 | * the group can be slid down. Similarly, if the last line in a group is equal |
| 684 | * to the line preceding the group, then the group can be slid up. See |
| 685 | * group_slide_down() and group_slide_up(). |
| 686 | * |
| 687 | * Note that loops that are testing for changed lines in xdf->rchg do not need |
| 688 | * index bounding since the array is prepared with a zero at position -1 and N. |
| 689 | */ |
| 690 | struct xdlgroup { |
| 691 | /* |
| 692 | * The index of the first changed line in the group, or the index of |
| 693 | * the unchanged line above which the (empty) group is located. |
| 694 | */ |
| 695 | long start; |
| 696 | |
| 697 | /* |
| 698 | * The index of the first unchanged line after the group. For an empty |
| 699 | * group, end is equal to start. |
| 700 | */ |
| 701 | long end; |
| 702 | }; |
| 703 | |
| 704 | /* |
| 705 | * Initialize g to point at the first group in xdf. |
| 706 | */ |
| 707 | static void group_init(xdfile_t *xdf, struct xdlgroup *g) |
| 708 | { |
| 709 | g->start = g->end = 0; |
| 710 | while (xdf->changed[g->end]) |
| 711 | g->end++; |
| 712 | } |
| 713 | |
| 714 | /* |
| 715 | * Move g to describe the next (possibly empty) group in xdf and return 0. If g |
| 716 | * is already at the end of the file, do nothing and return -1. |
| 717 | */ |
| 718 | static inline int group_next(xdfile_t *xdf, struct xdlgroup *g) |
| 719 | { |
| 720 | if (g->end == (long)xdf->nrec) |
| 721 | return -1; |
| 722 | |
| 723 | g->start = g->end + 1; |
| 724 | for (g->end = g->start; xdf->changed[g->end]; g->end++) |
| 725 | ; |
| 726 | |
| 727 | return 0; |
| 728 | } |
| 729 | |
| 730 | /* |
| 731 | * Move g to describe the previous (possibly empty) group in xdf and return 0. |
| 732 | * If g is already at the beginning of the file, do nothing and return -1. |
| 733 | */ |
| 734 | static inline int group_previous(xdfile_t *xdf, struct xdlgroup *g) |
| 735 | { |
| 736 | if (g->start == 0) |
| 737 | return -1; |
| 738 | |
| 739 | g->end = g->start - 1; |
| 740 | for (g->start = g->end; xdf->changed[g->start - 1]; g->start--) |
| 741 | ; |
| 742 | |
| 743 | return 0; |
| 744 | } |
| 745 | |
| 746 | /* |
| 747 | * If g can be slid toward the end of the file, do so, and if it bumps into a |
| 748 | * following group, expand this group to include it. Return 0 on success or -1 |
| 749 | * if g cannot be slid down. |
| 750 | */ |
| 751 | static int group_slide_down(xdfile_t *xdf, struct xdlgroup *g) |
| 752 | { |
| 753 | if (g->end < (long)xdf->nrec && |
| 754 | recs_match(&xdf->recs[g->start], &xdf->recs[g->end])) { |
| 755 | xdf->changed[g->start++] = false; |
| 756 | xdf->changed[g->end++] = true; |
| 757 | |
| 758 | while (xdf->changed[g->end]) |
| 759 | g->end++; |
| 760 | |
| 761 | return 0; |
| 762 | } else { |
| 763 | return -1; |
| 764 | } |
| 765 | } |
| 766 | |
| 767 | /* |
| 768 | * If g can be slid toward the beginning of the file, do so, and if it bumps |
| 769 | * into a previous group, expand this group to include it. Return 0 on success |
| 770 | * or -1 if g cannot be slid up. |
| 771 | */ |
| 772 | static int group_slide_up(xdfile_t *xdf, struct xdlgroup *g) |
| 773 | { |
| 774 | if (g->start > 0 && |
| 775 | recs_match(&xdf->recs[g->start - 1], &xdf->recs[g->end - 1])) { |
| 776 | xdf->changed[--g->start] = true; |
| 777 | xdf->changed[--g->end] = false; |
| 778 | |
| 779 | while (xdf->changed[g->start - 1]) |
| 780 | g->start--; |
| 781 | |
| 782 | return 0; |
| 783 | } else { |
| 784 | return -1; |
| 785 | } |
| 786 | } |
| 787 | |
| 788 | /* |
| 789 | * Move back and forward change groups for a consistent and pretty diff output. |
| 790 | * This also helps in finding joinable change groups and reducing the diff |
| 791 | * size. |
| 792 | */ |
| 793 | int xdl_change_compact(xdfile_t *xdf, xdfile_t *xdfo, long flags) { |
| 794 | struct xdlgroup g, go; |
| 795 | struct xdlgroup g_orig; |
| 796 | long earliest_end, end_matching_other; |
| 797 | long groupsize; |
| 798 | |
| 799 | group_init(xdf, &g); |
| 800 | group_init(xdfo, &go); |
| 801 | |
| 802 | while (1) { |
| 803 | /* |
| 804 | * If the group is empty in the to-be-compacted file, skip it: |
| 805 | */ |
| 806 | if (g.end == g.start) |
| 807 | goto next; |
| 808 | |
| 809 | g_orig = g; |
| 810 | |
| 811 | /* |
| 812 | * Now shift the change up and then down as far as possible in |
| 813 | * each direction. If it bumps into any other changes, merge |
| 814 | * them and restart the process. |
| 815 | */ |
| 816 | do { |
| 817 | groupsize = g.end - g.start; |
| 818 | |
| 819 | /* |
| 820 | * Keep track of the last "end" index that causes this |
| 821 | * group to align with a group of changed lines in the |
| 822 | * other file. -1 indicates that we haven't found such |
| 823 | * a match yet: |
| 824 | */ |
| 825 | end_matching_other = -1; |
| 826 | |
| 827 | /* Shift the group backward as much as possible: */ |
| 828 | while (!group_slide_up(xdf, &g)) |
| 829 | if (group_previous(xdfo, &go)) |
| 830 | BUG("group sync broken sliding up"); |
| 831 | |
| 832 | /* |
| 833 | * This is this highest that this group can be shifted. |
| 834 | * Record its end index: |
| 835 | */ |
| 836 | earliest_end = g.end; |
| 837 | |
| 838 | if (go.end > go.start) |
| 839 | end_matching_other = g.end; |
| 840 | |
| 841 | /* Now shift the group forward as far as possible: */ |
| 842 | while (1) { |
| 843 | if (group_slide_down(xdf, &g)) |
| 844 | break; |
| 845 | if (group_next(xdfo, &go)) |
| 846 | BUG("group sync broken sliding down"); |
| 847 | |
| 848 | if (go.end > go.start) |
| 849 | end_matching_other = g.end; |
| 850 | } |
| 851 | } while (groupsize != g.end - g.start); |
| 852 | |
| 853 | /* |
| 854 | * If the group can be shifted, then we can possibly use this |
| 855 | * freedom to produce a more intuitive diff. |
| 856 | * |
| 857 | * The group is currently shifted as far down as possible, so |
| 858 | * the heuristics below only have to handle upwards shifts. |
| 859 | */ |
| 860 | |
| 861 | if (g.end == earliest_end) { |
| 862 | /* no shifting was possible */ |
| 863 | } else if (end_matching_other != -1) { |
| 864 | /* |
| 865 | * Move the possibly merged group of changes back to |
| 866 | * line up with the last group of changes from the |
| 867 | * other file that it can align with. This avoids breaking |
| 868 | * a single change into a separate addition/deletion. |
| 869 | */ |
| 870 | while (go.end == go.start) { |
| 871 | if (group_slide_up(xdf, &g)) |
| 872 | BUG("match disappeared"); |
| 873 | if (group_previous(xdfo, &go)) |
| 874 | BUG("group sync broken sliding to match"); |
| 875 | } |
| 876 | } else if (flags & XDF_INDENT_HEURISTIC) { |
| 877 | /* |
| 878 | * Indent heuristic: a group of pure add/delete lines |
| 879 | * implies two splits, one between the end of the |
| 880 | * "before" context and the start of the group, and |
| 881 | * another between the end of the group and the |
| 882 | * beginning of the "after" context. Some splits are |
| 883 | * aesthetically better and some are worse. We compute |
| 884 | * a badness "score" for each split, and add the scores |
| 885 | * for the two splits to define a "score" for each |
| 886 | * position that the group can be shifted to. Then we |
| 887 | * pick the shift with the lowest score. |
| 888 | */ |
| 889 | long shift, best_shift = -1; |
| 890 | struct split_score best_score; |
| 891 | |
| 892 | shift = earliest_end; |
| 893 | if (g.end - groupsize - 1 > shift) |
| 894 | shift = g.end - groupsize - 1; |
| 895 | if (g.end - INDENT_HEURISTIC_MAX_SLIDING > shift) |
| 896 | shift = g.end - INDENT_HEURISTIC_MAX_SLIDING; |
| 897 | for (; shift <= g.end; shift++) { |
| 898 | struct split_measurement m; |
| 899 | struct split_score score = {0, 0}; |
| 900 | |
| 901 | measure_split(xdf, shift, &m); |
| 902 | score_add_split(&m, &score); |
| 903 | measure_split(xdf, shift - groupsize, &m); |
| 904 | score_add_split(&m, &score); |
| 905 | if (best_shift == -1 || |
| 906 | score_cmp(&score, &best_score) <= 0) { |
| 907 | best_score.effective_indent = score.effective_indent; |
| 908 | best_score.penalty = score.penalty; |
| 909 | best_shift = shift; |
| 910 | } |
| 911 | } |
| 912 | |
| 913 | while (g.end > best_shift) { |
| 914 | if (group_slide_up(xdf, &g)) |
| 915 | BUG("best shift unreached"); |
| 916 | if (group_previous(xdfo, &go)) |
| 917 | BUG("group sync broken sliding to blank line"); |
| 918 | } |
| 919 | } |
| 920 | |
| 921 | /* |
| 922 | * If we merged change groups during shifting, the new |
| 923 | * combined group could now have matching lines in both files, |
| 924 | * even if the original separate groups did not. Re-diff the |
| 925 | * new group to find these matching lines to mark them as |
| 926 | * unchanged. |
| 927 | * |
| 928 | * Only do this if the corresponding group in the other file is |
| 929 | * non-empty, as it's trivial otherwise. |
| 930 | * |
| 931 | * Only do this for histogram diff as its LCS algorithm allows |
| 932 | * for this scenario. In contrast, patience diff finds LCS |
| 933 | * of unique lines that groups cannot be shifted across. |
| 934 | * Myer's diff (standalone or used as fall-back in patience |
| 935 | * diff) already finds minimal edits so it is not possible for |
| 936 | * shifted groups to result in a smaller diff. (Without |
| 937 | * XDF_NEED_MINIMAL, Myer's isn't technically guaranteed to be |
| 938 | * minimal, but it should be so most of the time) |
| 939 | */ |
| 940 | if (go.end != go.start && |
| 941 | XDF_DIFF_ALG(flags) == XDF_HISTOGRAM_DIFF && |
| 942 | (g.start != g_orig.start || |
| 943 | g.end != g_orig.end)) { |
| 944 | xpparam_t xpp; |
| 945 | xdfenv_t xe; |
| 946 | |
| 947 | memset(&xpp, 0, sizeof(xpp)); |
| 948 | xpp.flags = flags & ~XDF_DIFF_ALGORITHM_MASK; |
| 949 | |
| 950 | xe.xdf1 = *xdf; |
| 951 | xe.xdf2 = *xdfo; |
| 952 | |
| 953 | if (xdl_fall_back_diff(&xe, &xpp, |
| 954 | g.start + 1, g.end - g.start, |
| 955 | go.start + 1, go.end - go.start)) { |
| 956 | return -1; |
| 957 | } |
| 958 | } |
| 959 | |
| 960 | next: |
| 961 | /* Move past the just-processed group: */ |
| 962 | if (group_next(xdf, &g)) |
| 963 | break; |
| 964 | if (group_next(xdfo, &go)) |
| 965 | BUG("group sync broken moving to next group"); |
| 966 | } |
| 967 | |
| 968 | if (!group_next(xdfo, &go)) |
| 969 | BUG("group sync broken at end of file"); |
| 970 | |
| 971 | return 0; |
| 972 | } |
| 973 | |
| 974 | |
| 975 | int xdl_build_script(xdfenv_t *xe, xdchange_t **xscr) { |
| 976 | xdchange_t *cscr = NULL, *xch; |
| 977 | bool *changed1 = xe->xdf1.changed, *changed2 = xe->xdf2.changed; |
| 978 | long i1, i2, l1, l2; |
| 979 | |
| 980 | /* |
| 981 | * Trivial. Collects "groups" of changes and creates an edit script. |
| 982 | */ |
| 983 | for (i1 = xe->xdf1.nrec, i2 = xe->xdf2.nrec; i1 >= 0 || i2 >= 0; i1--, i2--) |
| 984 | if (changed1[i1 - 1] || changed2[i2 - 1]) { |
| 985 | for (l1 = i1; changed1[i1 - 1]; i1--); |
| 986 | for (l2 = i2; changed2[i2 - 1]; i2--); |
| 987 | |
| 988 | if (!(xch = xdl_add_change(cscr, i1, i2, l1 - i1, l2 - i2))) { |
| 989 | xdl_free_script(cscr); |
| 990 | return -1; |
| 991 | } |
| 992 | cscr = xch; |
| 993 | } |
| 994 | |
| 995 | *xscr = cscr; |
| 996 | |
| 997 | return 0; |
| 998 | } |
| 999 | |
| 1000 | |
| 1001 | void xdl_free_script(xdchange_t *xscr) { |
| 1002 | xdchange_t *xch; |
| 1003 | |
| 1004 | while ((xch = xscr) != NULL) { |
| 1005 | xscr = xscr->next; |
| 1006 | xdl_free(xch); |
| 1007 | } |
| 1008 | } |
| 1009 | |
| 1010 | static int xdl_call_hunk_func(xdfenv_t *xe UNUSED, xdchange_t *xscr, xdemitcb_t *ecb, |
| 1011 | xdemitconf_t const *xecfg) |
| 1012 | { |
| 1013 | xdchange_t *xch, *xche; |
| 1014 | |
| 1015 | for (xch = xscr; xch; xch = xche->next) { |
| 1016 | xche = xdl_get_hunk(&xch, xecfg); |
| 1017 | if (!xch) |
| 1018 | break; |
| 1019 | if (xecfg->hunk_func(xch->i1, xche->i1 + xche->chg1 - xch->i1, |
| 1020 | xch->i2, xche->i2 + xche->chg2 - xch->i2, |
| 1021 | ecb->priv) < 0) |
| 1022 | return -1; |
| 1023 | } |
| 1024 | return 0; |
| 1025 | } |
| 1026 | |
| 1027 | static void xdl_mark_ignorable_lines(xdchange_t *xscr, xdfenv_t *xe, long flags) |
| 1028 | { |
| 1029 | xdchange_t *xch; |
| 1030 | |
| 1031 | for (xch = xscr; xch; xch = xch->next) { |
| 1032 | int ignore = 1; |
| 1033 | xrecord_t *rec; |
| 1034 | long i; |
| 1035 | |
| 1036 | rec = &xe->xdf1.recs[xch->i1]; |
| 1037 | for (i = 0; i < xch->chg1 && ignore; i++) |
| 1038 | ignore = xdl_blankline((const char *)rec[i].ptr, (long)rec[i].size, flags); |
| 1039 | |
| 1040 | rec = &xe->xdf2.recs[xch->i2]; |
| 1041 | for (i = 0; i < xch->chg2 && ignore; i++) |
| 1042 | ignore = xdl_blankline((const char *)rec[i].ptr, (long)rec[i].size, flags); |
| 1043 | |
| 1044 | xch->ignore = ignore; |
| 1045 | } |
| 1046 | } |
| 1047 | |
| 1048 | static int record_matches_regex(xrecord_t *rec, xpparam_t const *xpp) { |
| 1049 | regmatch_t regmatch; |
| 1050 | size_t i; |
| 1051 | |
| 1052 | for (i = 0; i < xpp->ignore_regex_nr; i++) |
| 1053 | if (!regexec_buf(xpp->ignore_regex[i], (const char *)rec->ptr, rec->size, 1, |
| 1054 | ®match, 0)) |
| 1055 | return 1; |
| 1056 | |
| 1057 | return 0; |
| 1058 | } |
| 1059 | |
| 1060 | static void xdl_mark_ignorable_regex(xdchange_t *xscr, const xdfenv_t *xe, |
| 1061 | xpparam_t const *xpp) |
| 1062 | { |
| 1063 | xdchange_t *xch; |
| 1064 | |
| 1065 | for (xch = xscr; xch; xch = xch->next) { |
| 1066 | xrecord_t *rec; |
| 1067 | int ignore = 1; |
| 1068 | long i; |
| 1069 | |
| 1070 | /* |
| 1071 | * Do not override --ignore-blank-lines. |
| 1072 | */ |
| 1073 | if (xch->ignore) |
| 1074 | continue; |
| 1075 | |
| 1076 | rec = &xe->xdf1.recs[xch->i1]; |
| 1077 | for (i = 0; i < xch->chg1 && ignore; i++) |
| 1078 | ignore = record_matches_regex(&rec[i], xpp); |
| 1079 | |
| 1080 | rec = &xe->xdf2.recs[xch->i2]; |
| 1081 | for (i = 0; i < xch->chg2 && ignore; i++) |
| 1082 | ignore = record_matches_regex(&rec[i], xpp); |
| 1083 | |
| 1084 | xch->ignore = ignore; |
| 1085 | } |
| 1086 | } |
| 1087 | |
| 1088 | int xdl_diff(mmfile_t *mf1, mmfile_t *mf2, xpparam_t const *xpp, |
| 1089 | xdemitconf_t const *xecfg, xdemitcb_t *ecb) { |
| 1090 | xdchange_t *xscr; |
| 1091 | xdfenv_t xe; |
| 1092 | emit_func_t ef = xecfg->hunk_func ? xdl_call_hunk_func : xdl_emit_diff; |
| 1093 | |
| 1094 | if (xdl_do_diff(mf1, mf2, xpp, &xe) < 0) { |
| 1095 | |
| 1096 | return -1; |
| 1097 | } |
| 1098 | if (xdl_change_compact(&xe.xdf1, &xe.xdf2, xpp->flags) < 0 || |
| 1099 | xdl_change_compact(&xe.xdf2, &xe.xdf1, xpp->flags) < 0 || |
| 1100 | xdl_build_script(&xe, &xscr) < 0) { |
| 1101 | |
| 1102 | xdl_free_env(&xe); |
| 1103 | return -1; |
| 1104 | } |
| 1105 | if (xscr) { |
| 1106 | if (xpp->flags & XDF_IGNORE_BLANK_LINES) |
| 1107 | xdl_mark_ignorable_lines(xscr, &xe, xpp->flags); |
| 1108 | |
| 1109 | if (xpp->ignore_regex) |
| 1110 | xdl_mark_ignorable_regex(xscr, &xe, xpp); |
| 1111 | |
| 1112 | if (ef(&xe, xscr, ecb, xecfg) < 0) { |
| 1113 | |
| 1114 | xdl_free_script(xscr); |
| 1115 | xdl_free_env(&xe); |
| 1116 | return -1; |
| 1117 | } |
| 1118 | xdl_free_script(xscr); |
| 1119 | } |
| 1120 | xdl_free_env(&xe); |
| 1121 | |
| 1122 | return 0; |
| 1123 | } |