| 1 | /* -*-c-*- |
| 2 | * |
| 3 | * Types for the test-vector framework |
| 4 | * |
| 5 | * (c) 2023 Straylight/Edgeware |
| 6 | */ |
| 7 | |
| 8 | /*----- Licensing notice --------------------------------------------------* |
| 9 | * |
| 10 | * This file is part of the mLib utilities library. |
| 11 | * |
| 12 | * mLib is free software: you can redistribute it and/or modify it under |
| 13 | * the terms of the GNU Library General Public License as published by |
| 14 | * the Free Software Foundation; either version 2 of the License, or (at |
| 15 | * your option) any later version. |
| 16 | * |
| 17 | * mLib is distributed in the hope that it will be useful, but WITHOUT |
| 18 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or |
| 19 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public |
| 20 | * License for more details. |
| 21 | * |
| 22 | * You should have received a copy of the GNU Library General Public |
| 23 | * License along with mLib. If not, write to the Free Software |
| 24 | * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, |
| 25 | * USA. |
| 26 | */ |
| 27 | |
| 28 | /*----- Header files ------------------------------------------------------*/ |
| 29 | |
| 30 | #include <assert.h> |
| 31 | #include <ctype.h> |
| 32 | #include <errno.h> |
| 33 | #include <float.h> |
| 34 | #include <limits.h> |
| 35 | #include <math.h> |
| 36 | #include <stdio.h> |
| 37 | #include <string.h> |
| 38 | |
| 39 | #include "buf.h" |
| 40 | #include "codec.h" |
| 41 | # include "base32.h" |
| 42 | # include "base64.h" |
| 43 | # include "hex.h" |
| 44 | #include "dstr.h" |
| 45 | #include "maths.h" |
| 46 | #include "tvec.h" |
| 47 | |
| 48 | /*----- Preliminary utilities ---------------------------------------------*/ |
| 49 | |
| 50 | /* --- @trivial_release@ --- * |
| 51 | * |
| 52 | * Arguments: @union tvec_regval *rv@ = a register value |
| 53 | * @const struct tvec_regdef@ = the register definition |
| 54 | * |
| 55 | * Returns: --- |
| 56 | * |
| 57 | * Use: Does nothing. Used for register values which don't retain |
| 58 | * resources. |
| 59 | */ |
| 60 | |
| 61 | static void trivial_release(union tvec_regval *rv, |
| 62 | const struct tvec_regdef *rd) |
| 63 | { ; } |
| 64 | |
| 65 | /*----- Integer utilities -------------------------------------------------*/ |
| 66 | |
| 67 | /* --- @unsigned_to_buf@, @signed_to_buf@ --- * |
| 68 | * |
| 69 | * Arguments: @buf *b@ = buffer to write on |
| 70 | * @unsigned long u@ or @long i@ = integer to write |
| 71 | * |
| 72 | * Returns: Zero on success, @-1@ on failure. |
| 73 | * |
| 74 | * Use: Write @i@ to the buffer, in big-endian (two's-complement, it |
| 75 | * signed) format. |
| 76 | */ |
| 77 | |
| 78 | static int unsigned_to_buf(buf *b, unsigned long u) |
| 79 | { kludge64 k; ASSIGN64(k, u); return (buf_putk64l(b, k)); } |
| 80 | |
| 81 | static int signed_to_buf(buf *b, long i) |
| 82 | { |
| 83 | kludge64 k; |
| 84 | unsigned long u; |
| 85 | |
| 86 | u = i; |
| 87 | if (i >= 0) ASSIGN64(k, u); |
| 88 | else { ASSIGN64(k, ~u); CPL64(k, k); } |
| 89 | return (buf_putk64l(b, k)); |
| 90 | } |
| 91 | |
| 92 | /* --- @unsigned_from_buf@, @signed_from_buf@ --- * |
| 93 | * |
| 94 | * Arguments: @buf *b@ = buffer to write on |
| 95 | * @unsigned long *u_out@ or @long *i_out@ = where to put the |
| 96 | * result |
| 97 | * |
| 98 | * Returns: Zero on success, @-1@ on failure. |
| 99 | * |
| 100 | * Use: Read an integer, in big-endian (two's-complement, if signed) |
| 101 | * format, from the buffer. |
| 102 | */ |
| 103 | |
| 104 | static int unsigned_from_buf(buf *b, unsigned long *u_out) |
| 105 | { |
| 106 | kludge64 k, ulmax; |
| 107 | |
| 108 | ASSIGN64(ulmax, ULONG_MAX); |
| 109 | if (buf_getk64l(b, &k)) return (-1); |
| 110 | if (CMP64(k, >, ulmax)) { buf_break(b); return (-1); } |
| 111 | *u_out = GET64(unsigned long, k); return (0); |
| 112 | } |
| 113 | |
| 114 | /* --- @hex_width@ --- * |
| 115 | * |
| 116 | * Arguments: @unsigned long u@ = an integer |
| 117 | * |
| 118 | * Returns: A suitable number of digits to use in order to display @u@ in |
| 119 | * hex. Currently, we select a power of two sufficient to show |
| 120 | * the value, but at least 2. |
| 121 | */ |
| 122 | |
| 123 | static int hex_width(unsigned long u) |
| 124 | { |
| 125 | int wd; |
| 126 | unsigned long t; |
| 127 | |
| 128 | for (t = u >> 4, wd = 4; t >>= wd, wd *= 2, t; ); |
| 129 | return (wd/4); |
| 130 | } |
| 131 | |
| 132 | /* --- @format_unsigned_hex@, @format_signed_hex@ --- * |
| 133 | * |
| 134 | * Arguments: @const struct gprintf_ops *gops@ = print operations |
| 135 | * @void *go@ = print destination |
| 136 | * @unsigned long u@ or @long i@ = integer to print |
| 137 | * |
| 138 | * Returns: --- |
| 139 | * |
| 140 | * Use: Print an unsigned or signed integer in hexadecimal. |
| 141 | */ |
| 142 | |
| 143 | static void format_unsigned_hex(const struct gprintf_ops *gops, void *go, |
| 144 | unsigned long u) |
| 145 | { gprintf(gops, go, "0x%0*lx", hex_width(u), u); } |
| 146 | |
| 147 | static void format_signed_hex(const struct gprintf_ops *gops, void *go, |
| 148 | long i) |
| 149 | { |
| 150 | unsigned long u = i >= 0 ? i : -(unsigned long)i; |
| 151 | gprintf(gops, go, "%s0x%0*lx", i < 0 ? "-" : "", hex_width(u), u); |
| 152 | } |
| 153 | |
| 154 | static int signed_from_buf(buf *b, long *i_out) |
| 155 | { |
| 156 | kludge64 k, lmax, not_lmin; |
| 157 | |
| 158 | ASSIGN64(lmax, LONG_MAX); ASSIGN64(not_lmin, ~(unsigned long)LONG_MIN); |
| 159 | if (buf_getk64l(b, &k)) return (-1); |
| 160 | if (CMP64(k, <=, lmax)) *i_out = (long)GET64(unsigned long, k); |
| 161 | else { |
| 162 | CPL64(k, k); |
| 163 | if (CMP64(k, <=, not_lmin)) *i_out = -(long)GET64(unsigned long, k) - 1; |
| 164 | else { buf_break(b); return (-1); } |
| 165 | } |
| 166 | return (0); |
| 167 | } |
| 168 | |
| 169 | /* --- @check_unsigned_range@, @check_signed_range@ --- * |
| 170 | * |
| 171 | * Arguments: @unsigned long u@ or @long i@ = an integer |
| 172 | * @const struct tvec_urange *ur@ or |
| 173 | * @const struct tvec_irange *ir@ = range specification, |
| 174 | * or null |
| 175 | * @struct tvec_state *tv@ = test vector state |
| 176 | * |
| 177 | * Returns: Zero on success, or @-1@ on error. |
| 178 | * |
| 179 | * Use: Check that the integer is within bounds. If not, report a |
| 180 | * suitable error and return a failure indication. |
| 181 | */ |
| 182 | |
| 183 | static int check_signed_range(long i, |
| 184 | const struct tvec_irange *ir, |
| 185 | struct tvec_state *tv) |
| 186 | { |
| 187 | if (ir && (ir->min > i || i > ir->max)) { |
| 188 | tvec_error(tv, "integer %ld out of range (must be in [%ld .. %ld])", |
| 189 | i, ir->min, ir->max); |
| 190 | return (-1); |
| 191 | } |
| 192 | return (0); |
| 193 | } |
| 194 | |
| 195 | static int check_unsigned_range(unsigned long u, |
| 196 | const struct tvec_urange *ur, |
| 197 | struct tvec_state *tv) |
| 198 | { |
| 199 | if (ur && (ur->min > u || u > ur->max)) { |
| 200 | tvec_error(tv, "integer %lu out of range (must be in [%lu .. %lu])", |
| 201 | u, ur->min, ur->max); |
| 202 | return (-1); |
| 203 | } |
| 204 | return (0); |
| 205 | } |
| 206 | |
| 207 | /* --- @chtodig@ --- * |
| 208 | * |
| 209 | * Arguments: @int ch@ = a character |
| 210 | * |
| 211 | * Returns: The numeric value of the character as a digit, or @-1@ if |
| 212 | * it's not a digit. Letters count as extended digits starting |
| 213 | * with value 10; case is not significant. |
| 214 | */ |
| 215 | |
| 216 | static int chtodig(int ch) |
| 217 | { |
| 218 | if ('0' <= ch && ch <= '9') return (ch - '0'); |
| 219 | else if ('a' <= ch && ch <= 'z') return (ch - 'a' + 10); |
| 220 | else if ('A' <= ch && ch <= 'Z') return (ch - 'A' + 10); |
| 221 | else return (-1); |
| 222 | } |
| 223 | |
| 224 | /* --- @parse_unsigned_integer@, @parse_signed_integer@ --- * |
| 225 | * |
| 226 | * Arguments: @unsigned long *u_out@, @long *i_out@ = where to put the |
| 227 | * result |
| 228 | * @const char **q_out@ = where to put the end position |
| 229 | * @const char *p@ = pointer to the string to parse |
| 230 | * |
| 231 | * Returns: Zero on success, @-1@ on error. |
| 232 | * |
| 233 | * Use: Parse an integer from a string in the test-vector format. |
| 234 | * This is mostly extension of the traditional C @strtoul@ |
| 235 | * format: supported inputs include: |
| 236 | * |
| 237 | * * NNN -- a decimal number (even if it starts with `0'); |
| 238 | * * 0xNNN -- hexadecimal; |
| 239 | * * 0oNNN -- octal; |
| 240 | * * 0bNNN -- binary; |
| 241 | * * NNrNNN -- base NN. |
| 242 | * |
| 243 | * Furthermore, single underscores are permitted internally as |
| 244 | * an insignificant digit separator. |
| 245 | */ |
| 246 | |
| 247 | static int parse_unsigned_integer(unsigned long *u_out, const char **q_out, |
| 248 | const char *p) |
| 249 | { |
| 250 | unsigned long u; |
| 251 | int ch, d, r; |
| 252 | const char *q; |
| 253 | unsigned f = 0; |
| 254 | #define f_implicit 1u /* implicitly reading base 10 */ |
| 255 | #define f_digit 2u /* read a real digit */ |
| 256 | #define f_uscore 4u /* found an underscore */ |
| 257 | |
| 258 | /* Initial setup |
| 259 | * |
| 260 | * This will deal with the traditional `0[box]...' prefixes. We'll leave |
| 261 | * our new `NNr...' syntax for later. |
| 262 | */ |
| 263 | if (p[0] != '0' || !p[1]) { |
| 264 | d = chtodig(*p); if (0 > d || d >= 10) return (-1); |
| 265 | r = 10; u = d; p++; f |= f_implicit | f_digit; |
| 266 | } else { |
| 267 | u = 0; d = chtodig(p[2]); |
| 268 | if (d < 0) { r = 10; f |= f_implicit | f_digit; p++; } |
| 269 | else if ((p[1] == 'x' || p[1] == 'X') && d < 16) { r = 16; p += 2; } |
| 270 | else if ((p[1] == 'o' || p[1] == 'O') && d < 8) { r = 8; p += 2; } |
| 271 | else if ((p[1] == 'b' || p[1] == 'B') && d < 2) { r = 2; p += 2; } |
| 272 | else { r = 10; f |= f_digit; p++; } |
| 273 | } |
| 274 | |
| 275 | q = p; |
| 276 | for (;;) { |
| 277 | /* Work through the string a character at a time. */ |
| 278 | |
| 279 | ch = *p; switch (ch) { |
| 280 | |
| 281 | case '_': |
| 282 | /* An underscore is OK if we haven't just seen one. */ |
| 283 | |
| 284 | if (f&f_uscore) goto done; |
| 285 | p++; f = (f&~f_implicit) | f_uscore; |
| 286 | break; |
| 287 | |
| 288 | case 'r': case 'R': |
| 289 | /* An `r' is OK if the number so far is small enough to be a sensible |
| 290 | * base, and we're scanning decimal implicitly. |
| 291 | */ |
| 292 | |
| 293 | if (!(f&f_implicit) || !u || u >= 36) goto done; |
| 294 | d = chtodig(p[1]); if (0 > d || d >= u) goto done; |
| 295 | r = u; u = d; f = (f&~f_implicit) | f_digit; p += 2; q = p; |
| 296 | break; |
| 297 | |
| 298 | default: |
| 299 | /* Otherwise we expect a valid digit and accumulate it. */ |
| 300 | d = chtodig(ch); if (d < 0 || d >= r) goto done; |
| 301 | if (u > ULONG_MAX/r) return (-1); |
| 302 | u *= r; if (u > ULONG_MAX - d) return (-1); |
| 303 | u += d; f = (f&~f_uscore) | f_digit; p++; q = p; |
| 304 | break; |
| 305 | } |
| 306 | } |
| 307 | |
| 308 | done: |
| 309 | if (!(f&f_digit)) return (-1); |
| 310 | *u_out = u; *q_out = q; return (0); |
| 311 | |
| 312 | #undef f_implicit |
| 313 | #undef f_digit |
| 314 | #undef f_uscore |
| 315 | } |
| 316 | |
| 317 | static int parse_signed_integer(long *i_out, const char **q_out, |
| 318 | const char *p) |
| 319 | { |
| 320 | unsigned long u; |
| 321 | unsigned f = 0; |
| 322 | #define f_neg 1u |
| 323 | |
| 324 | /* Read an initial sign. */ |
| 325 | if (*p == '+') p++; |
| 326 | else if (*p == '-') { f |= f_neg; p++; } |
| 327 | |
| 328 | /* Scan an unsigned number. */ |
| 329 | if (parse_unsigned_integer(&u, q_out, p)) return (-1); |
| 330 | |
| 331 | /* Check for signed overflow and apply the sign. */ |
| 332 | if (!(f&f_neg)) { |
| 333 | if (u > LONG_MAX) return (-1); |
| 334 | *i_out = u; |
| 335 | } else { |
| 336 | if (u && u - 1 > -(LONG_MIN + 1)) return (-1); |
| 337 | *i_out = u ? -(long)(u - 1) - 1 : 0; |
| 338 | } |
| 339 | |
| 340 | return (0); |
| 341 | |
| 342 | #undef f_neg |
| 343 | } |
| 344 | |
| 345 | /* --- @parse_unsigned@, @parse_signed@ --- * |
| 346 | * |
| 347 | * Arguments: @unsigned long *u_out@ or @long *i_out@ = where to put the |
| 348 | * result |
| 349 | * @const char *p@ = string to parse |
| 350 | * @const struct tvec_urange *ur@ or |
| 351 | * @const struct tvec_irange *ir@ = range specification, |
| 352 | * or null |
| 353 | * @struct tvec_state *tv@ = test vector state |
| 354 | * |
| 355 | * Returns: Zero on success, @-1@ on error. |
| 356 | * |
| 357 | * Use: Parse and range-check an integer. Unlike @parse_(un)signed_ |
| 358 | * integer@, these functions check that there's no cruft |
| 359 | * following the final digit, and report errors as they find |
| 360 | * them rather than leaving that to the caller. |
| 361 | */ |
| 362 | |
| 363 | static int parse_unsigned(unsigned long *u_out, const char *p, |
| 364 | const struct tvec_urange *ur, |
| 365 | struct tvec_state *tv) |
| 366 | { |
| 367 | unsigned long u; |
| 368 | const char *q; |
| 369 | |
| 370 | if (parse_unsigned_integer(&u, &q, p)) |
| 371 | return (tvec_error(tv, "invalid unsigned integer `%s'", p)); |
| 372 | if (*q) return (tvec_syntax(tv, *q, "end-of-line")); |
| 373 | if (check_unsigned_range(u, ur, tv)) return (-1); |
| 374 | *u_out = u; return (0); |
| 375 | } |
| 376 | |
| 377 | static int parse_signed(long *i_out, const char *p, |
| 378 | const struct tvec_irange *ir, |
| 379 | struct tvec_state *tv) |
| 380 | { |
| 381 | long i; |
| 382 | const char *q; |
| 383 | |
| 384 | if (parse_signed_integer(&i, &q, p)) |
| 385 | return (tvec_error(tv, "invalid signed integer `%s'", p)); |
| 386 | if (*q) return (tvec_syntax(tv, *q, "end-of-line")); |
| 387 | if (check_signed_range(i, ir, tv)) return (-1); |
| 388 | *i_out = i; return (0); |
| 389 | } |
| 390 | static const char size_units[] = "kMGTPEZY"; |
| 391 | |
| 392 | /* --- @parse_size@ --- * |
| 393 | * |
| 394 | * Arguments: @struct tvec_state *tv@ = test-vector state |
| 395 | * @size_t *u_out@ = where to put the answer |
| 396 | * @const char *delims@ = delimiters |
| 397 | * @const char *what@ = description of what we're parsing |
| 398 | * |
| 399 | * Returns: Zero on success, %$-1$% on failure. |
| 400 | * |
| 401 | * Use: Parse a memory size. |
| 402 | */ |
| 403 | |
| 404 | static int parse_size(struct tvec_state *tv, size_t *u_out, |
| 405 | const char *delims, const char *what) |
| 406 | { |
| 407 | dstr d = DSTR_INIT; |
| 408 | const char *p, *unit; |
| 409 | unsigned long u, t; |
| 410 | int rc; |
| 411 | unsigned f = 0; |
| 412 | #define f_range 1u |
| 413 | |
| 414 | if (tvec_readword(tv, &d, 0, delims, what)) { rc = -1; goto end; } |
| 415 | p = d.buf; |
| 416 | if (parse_unsigned_integer(&u, &p, p)) goto bad; |
| 417 | if (!*p) tvec_readword(tv, &d, &p, delims, 0); |
| 418 | |
| 419 | if (u > (size_t)-1) goto rangerr; |
| 420 | for (t = u, unit = size_units; *unit; unit++) { |
| 421 | if (t > (size_t)-1/1024) f |= f_range; |
| 422 | else t *= 1024; |
| 423 | if (*p == *unit) { |
| 424 | if (f&f_range) goto rangerr; |
| 425 | u = t; p++; break; |
| 426 | } |
| 427 | } |
| 428 | if (*p == 'B') p++; |
| 429 | if (*p) goto bad; |
| 430 | |
| 431 | *u_out = u; rc = 0; |
| 432 | end: |
| 433 | dstr_destroy(&d); |
| 434 | return (rc); |
| 435 | |
| 436 | bad: |
| 437 | tvec_error(tv, "invalid %s `%s'", what, d.buf); |
| 438 | rc = -1; goto end; |
| 439 | |
| 440 | rangerr: |
| 441 | tvec_error(tv, "%s `%s' out of range", what, d.buf); |
| 442 | rc = -1; goto end; |
| 443 | |
| 444 | #undef f_range |
| 445 | } |
| 446 | |
| 447 | /* --- @format_size@ --- * |
| 448 | * |
| 449 | * Arguments: @const struct gprintf_ops *gops@ = print operations |
| 450 | * @void *go@ = print destination |
| 451 | * @unsigned long u@ = a size |
| 452 | * @unsigned style@ = style (@TVSF_...@) |
| 453 | * |
| 454 | * Returns: --- |
| 455 | * |
| 456 | * Use: Format @u@ as a size in bytes to the destination, expressing |
| 457 | * it with a unit prefix if this is possible exactly. |
| 458 | */ |
| 459 | |
| 460 | static void format_size(const struct gprintf_ops *gops, void *go, |
| 461 | unsigned long u, unsigned style) |
| 462 | { |
| 463 | const char *unit; |
| 464 | |
| 465 | if (!u || u%1024) |
| 466 | gprintf(gops, go, "%lu%sB", u, style&TVSF_COMPACT ? "" : " "); |
| 467 | else { |
| 468 | for (unit = size_units, u /= 1024; |
| 469 | !(u%1024) && unit[1]; |
| 470 | u /= 1024, unit++); |
| 471 | gprintf(gops, go, "%lu%s%cB", u, style&TVSF_COMPACT ? "" : " ", *unit); |
| 472 | } |
| 473 | } |
| 474 | |
| 475 | /*----- Floating-point utilities ------------------------------------------*/ |
| 476 | |
| 477 | /* --- @eqish_floating_p@ --- * |
| 478 | * |
| 479 | * Arguments: @double x, y@ = two numbers to compare |
| 480 | * @const struct tvec_floatinfo *fi@ = floating-point info |
| 481 | * |
| 482 | * Returns: Nonzero if the comparand @y@ is sufficiently close to the |
| 483 | * reference @x@, or zero if it's definitely different. |
| 484 | */ |
| 485 | |
| 486 | static int eqish_floating_p(double x, double y, |
| 487 | const struct tvec_floatinfo *fi) |
| 488 | { |
| 489 | double t; |
| 490 | |
| 491 | if (NANP(x)) return (NANP(y)); else if (NANP(y)) return (0); |
| 492 | if (INFP(x)) return (x == y); else if (INFP(y)) return (0); |
| 493 | |
| 494 | switch (fi ? fi->f&TVFF_EQMASK : TVFF_EXACT) { |
| 495 | case TVFF_EXACT: |
| 496 | return (x == y && NEGP(x) == NEGP(y)); |
| 497 | case TVFF_ABSDELTA: |
| 498 | t = x - y; if (t < 0) t = -t; return (t < fi->delta); |
| 499 | case TVFF_RELDELTA: |
| 500 | t = 1.0 - y/x; if (t < 0) t = -t; return (t < fi->delta); |
| 501 | default: |
| 502 | abort(); |
| 503 | } |
| 504 | } |
| 505 | |
| 506 | /* --- @format_floating@ --- * |
| 507 | * |
| 508 | * Arguments: @const struct gprintf_ops *gops@ = print operations |
| 509 | * @void *go@ = print destination |
| 510 | * @double x@ = number to print |
| 511 | * |
| 512 | * Returns: --- |
| 513 | * |
| 514 | * Use: Print a floating-point number, accurately. |
| 515 | */ |
| 516 | |
| 517 | static void format_floating(const struct gprintf_ops *gops, void *go, |
| 518 | double x) |
| 519 | { |
| 520 | int prec; |
| 521 | |
| 522 | if (NANP(x)) |
| 523 | gprintf(gops, go, "#nan"); |
| 524 | else if (INFP(x)) |
| 525 | gprintf(gops, go, x > 0 ? "#+inf" : "#-inf"); |
| 526 | else { |
| 527 | /* Ugh. C doesn't provide any function for just printing a |
| 528 | * floating-point number /correctly/, i.e., so that you can read the |
| 529 | * result back and recover the number you first thought of. There are |
| 530 | * complicated algorithms published for doing this, but I really don't |
| 531 | * want to get into that here. So we have this. |
| 532 | * |
| 533 | * The sign doesn't cause significant difficulty so we're going to ignore |
| 534 | * it for now. So suppose we're given a number %$x = f b^e$%, in |
| 535 | * base-%$b$% format, so %$f b^n$% and %$e$% are integers, with |
| 536 | * %$0 \le f < 1$%. We're going to convert it into the nearest integer |
| 537 | * of the form %$X = F B^E$%, with similar conditions, only with the |
| 538 | * additional requirement that %$X$% is normalized, i.e., that %$X = 0$% |
| 539 | * or %$F \ge B^{-N}$%. |
| 540 | * |
| 541 | * We're rounding to the nearest such %$X$%. If there is to be ambiguity |
| 542 | * in the conversion, then some %$x = f b^e$% and the next smallest |
| 543 | * representable number %$x' = x + b^{e-n}$% must both map to the same |
| 544 | * %$X$%, which means both %$x$% and %$x'$% must be nearer to %$X$% than |
| 545 | * any other number representable in the target system. The nest larger |
| 546 | * number is %$X' = X + B^{E-N}$%; the next smaller number will normally |
| 547 | * be %$W = X - B^{E-N}$%, but if %$F = 1/B$ then the next smaller number |
| 548 | * is actually %$X - B^{E-N-1}$%. We ignore this latter possibility in |
| 549 | * the pursuit of a conservative estimate (though actually it doesn't |
| 550 | * matter). |
| 551 | * |
| 552 | * If both %$x$% and %$x'$% map to %$X$% then we must have |
| 553 | * %$L = X - B^{E-N}/2 \le x$% and %$x + b^{e-n} \le R = X + B^{E-N}/2$%; |
| 554 | * so firstly %$f b^e = x \ge L = W + B^{E-N}/2 > W = (F - B^{-N}) B^E$%, |
| 555 | * and secondly %$b^{e-n} \le B^{E-N}$%. Since these inequalities are in |
| 556 | * opposite senses, we can divide, giving |
| 557 | * |
| 558 | * %$f b^e/b^{e-n} > (F - B^{-N}) B^E/B^{E-N}$% , |
| 559 | * |
| 560 | * whence |
| 561 | * |
| 562 | * %$f b^n > (F - B^{-N}) B^N = F B^N - 1$% . |
| 563 | * |
| 564 | * Now %$f \le 1 - b^{-n}$%, and %$F \ge B^{-1}$%, so, for this to be |
| 565 | * possible, it must be the case that |
| 566 | * |
| 567 | * %$(1 - b^{-n}) b^n = b^n - 1 > B^{N-1} - 1$% . |
| 568 | * |
| 569 | * Then rearrange and take logarithms, obtaining |
| 570 | * |
| 571 | * %$(N - 1) \log B < n \log b$% , |
| 572 | * |
| 573 | * and so |
| 574 | * |
| 575 | * %$N < n \log b/\log B + 1$% . |
| 576 | * |
| 577 | * Recall that this is a necessary condition for a collision to occur; we |
| 578 | * are therefore safe whenever |
| 579 | * |
| 580 | * %$N \ge n \log b/\log B + 1$% ; |
| 581 | * |
| 582 | * so, taking ceilings, |
| 583 | * |
| 584 | * %$N \ge \lceil n \log b/\log B \rceil + 1$% . |
| 585 | * |
| 586 | * So that's why we have this. |
| 587 | * |
| 588 | * I'm going to assume that @n = DBL_MANT_DIG@ is sufficiently small that |
| 589 | * we can calculate this without ending up on the wrong side of an |
| 590 | * integer boundary. |
| 591 | * |
| 592 | * In C11, we have @DBL_DECIMAL_DIG@, which should be the same value only |
| 593 | * as a constant. Except that modern compilers are more than clever |
| 594 | * enough to work out that this is a constant anyway. |
| 595 | * |
| 596 | * This is sometimes an overestimate: we'll print out meaningless digits |
| 597 | * that don't represent anything we actually know about the number in |
| 598 | * question. To fix that, we'd need a complicated algorithm like Steele |
| 599 | * and White's Dragon4, Gay's @dtoa@, or Burger and Dybvig's algorithm |
| 600 | * (note that Loitsch's Grisu2 is conservative, and Grisu3 hands off to |
| 601 | * something else in difficult situations). |
| 602 | */ |
| 603 | |
| 604 | prec = ceil(DBL_MANT_DIG*log(FLT_RADIX)/log(10)) + 1; |
| 605 | gprintf(gops, go, "%.*g", prec, x); |
| 606 | } |
| 607 | } |
| 608 | |
| 609 | /* --- @parse_floating@ --- * |
| 610 | * |
| 611 | * Arguments: @double *x_out@ = where to put the result |
| 612 | * @const char *q_out@ = where to leave end pointer, or null |
| 613 | * @const char *p@ = string to parse |
| 614 | * @const struct tvec_floatinfo *fi@ = floating-point info |
| 615 | * @struct tvec_state *tv@ = test vector state |
| 616 | * |
| 617 | * Returns: Zero on success, @-1@ on error. |
| 618 | * |
| 619 | * Use: Parse a floating-point number from a string. Reports any |
| 620 | * necessary errors. If @q_out@ is not null then trailing |
| 621 | * material is permitted and a pointer to it (or the end of the |
| 622 | * string) is left in @*q_out@. |
| 623 | */ |
| 624 | |
| 625 | static int parse_floating(double *x_out, const char **q_out, const char *p, |
| 626 | const struct tvec_floatinfo *fi, |
| 627 | struct tvec_state *tv) |
| 628 | { |
| 629 | const char *pp; char *q; |
| 630 | dstr d = DSTR_INIT; |
| 631 | double x; |
| 632 | int olderr, rc; |
| 633 | |
| 634 | /* Check for special tokens. */ |
| 635 | if (STRCMP(p, ==, "#nan")) { |
| 636 | #ifdef NAN |
| 637 | if (q_out) *q_out = p + strlen(p); |
| 638 | x = NAN; rc = 0; |
| 639 | #else |
| 640 | tvec_error(tv, "NaN not supported on this system"); |
| 641 | rc = -1; goto end; |
| 642 | #endif |
| 643 | } |
| 644 | |
| 645 | else if (STRCMP(p, ==, "#inf") || |
| 646 | STRCMP(p, ==, "#+inf") || STRCMP(p, ==, "+#inf")) { |
| 647 | #ifdef INFINITY |
| 648 | if (q_out) *q_out = p + strlen(p); |
| 649 | x = INFINITY; rc = 0; |
| 650 | #else |
| 651 | tvec_error(tv, "infinity not supported on this system"); |
| 652 | rc = -1; goto end; |
| 653 | #endif |
| 654 | } |
| 655 | |
| 656 | else if (STRCMP(p, ==, "#-inf") || STRCMP(p, ==, "-#inf")) { |
| 657 | #ifdef INFINITY |
| 658 | if (q_out) *q_out = p + strlen(p); |
| 659 | x = -INFINITY; rc = 0; |
| 660 | #else |
| 661 | tvec_error(tv, "infinity not supported on this system"); |
| 662 | rc = -1; goto end; |
| 663 | #endif |
| 664 | } |
| 665 | |
| 666 | /* Check that this looks like a number, so we can exclude `strtod' |
| 667 | * recognizing its own non-finite number tokens. |
| 668 | */ |
| 669 | else { |
| 670 | pp = p; |
| 671 | if (*pp == '+' || *pp == '-') pp++; |
| 672 | if (*pp == '.') pp++; |
| 673 | if (!ISDIGIT(*pp)) { |
| 674 | tvec_syntax(tv, *p ? *p : fgetc(tv->fp), "floating-point number"); |
| 675 | rc = -1; goto end; |
| 676 | } |
| 677 | |
| 678 | /* Parse the number using the system parser. */ |
| 679 | olderr = errno; errno = 0; |
| 680 | x = strtod(p, &q); |
| 681 | if (q_out) *q_out = q; |
| 682 | else if (*q) { tvec_syntax(tv, *q, "end-of-line"); rc = -1; goto end; } |
| 683 | if (errno && (errno != ERANGE || (x > 0 ? -x : x) == HUGE_VAL)) { |
| 684 | tvec_error(tv, "invalid floating-point number `%.*s': %s", |
| 685 | (int)(q - p), p, strerror(errno)); |
| 686 | rc = -1; goto end; |
| 687 | } |
| 688 | errno = olderr; |
| 689 | } |
| 690 | |
| 691 | /* Check that the number is acceptable. */ |
| 692 | if (NANP(x) && fi && !(fi->f&TVFF_NANOK)) { |
| 693 | tvec_error(tv, "#nan not allowed here"); |
| 694 | rc = -1; goto end; |
| 695 | } |
| 696 | |
| 697 | if (fi && ((!(fi->f&TVFF_NOMIN) && x < fi->min) || |
| 698 | (!(fi->f&TVFF_NOMAX) && x > fi->max))) { |
| 699 | dstr_puts(&d, "floating-point number "); |
| 700 | format_floating(&dstr_printops, &d, x); |
| 701 | dstr_puts(&d, " out of range (must be in "); |
| 702 | if (fi->f&TVFF_NOMIN) |
| 703 | dstr_puts(&d, "(#-inf"); |
| 704 | else |
| 705 | { dstr_putc(&d, '['); format_floating(&dstr_printops, &d, fi->min); } |
| 706 | dstr_puts(&d, " .. "); |
| 707 | if (fi->f&TVFF_NOMAX) |
| 708 | dstr_puts(&d, "#+inf)"); |
| 709 | else |
| 710 | { format_floating(&dstr_printops, &d, fi->max); dstr_putc(&d, ']'); } |
| 711 | dstr_putc(&d, ')'); dstr_putz(&d); |
| 712 | tvec_error(tv, "%s", d.buf); rc = -1; goto end; |
| 713 | } |
| 714 | |
| 715 | /* All done. */ |
| 716 | *x_out = x; rc = 0; |
| 717 | end: |
| 718 | dstr_destroy(&d); |
| 719 | return (rc); |
| 720 | } |
| 721 | |
| 722 | /*----- String utilities --------------------------------------------------*/ |
| 723 | |
| 724 | /* Special character name table. */ |
| 725 | static const struct chartab { |
| 726 | const char *name; /* character name */ |
| 727 | int ch; /* character value */ |
| 728 | unsigned f; /* flags: */ |
| 729 | #define CTF_PREFER 1u /* preferred name */ |
| 730 | #define CTF_SHORT 2u /* short name (compact style) */ |
| 731 | } chartab[] = { |
| 732 | { "#eof", EOF, CTF_PREFER | CTF_SHORT }, |
| 733 | { "#nul", '\0', CTF_PREFER }, |
| 734 | { "#bell", '\a', CTF_PREFER }, |
| 735 | { "#ding", '\a', 0 }, |
| 736 | { "#bel", '\a', CTF_SHORT }, |
| 737 | { "#backspace", '\b', CTF_PREFER }, |
| 738 | { "#bs", '\b', CTF_SHORT }, |
| 739 | { "#escape", '\x1b', CTF_PREFER }, |
| 740 | { "#esc", '\x1b', CTF_SHORT }, |
| 741 | { "#formfeed", '\f', CTF_PREFER }, |
| 742 | { "#ff", '\f', CTF_SHORT }, |
| 743 | { "#newline", '\n', CTF_PREFER }, |
| 744 | { "#linefeed", '\n', 0 }, |
| 745 | { "#lf", '\n', CTF_SHORT }, |
| 746 | { "#nl", '\n', 0 }, |
| 747 | { "#return", '\r', CTF_PREFER }, |
| 748 | { "#carriage-return", '\r', 0 }, |
| 749 | { "#cr", '\r', CTF_SHORT }, |
| 750 | { "#tab", '\t', CTF_PREFER | CTF_SHORT }, |
| 751 | { "#horizontal-tab", '\t', 0 }, |
| 752 | { "#ht", '\t', 0 }, |
| 753 | { "#vertical-tab", '\v', CTF_PREFER }, |
| 754 | { "#vt", '\v', CTF_SHORT }, |
| 755 | { "#space", ' ', 0 }, |
| 756 | { "#spc", ' ', CTF_SHORT }, |
| 757 | { "#delete", '\x7f', CTF_PREFER }, |
| 758 | { "#del", '\x7f', CTF_SHORT }, |
| 759 | { 0, 0, 0 } |
| 760 | }; |
| 761 | |
| 762 | /* --- @find_charname@ --- * |
| 763 | * |
| 764 | * Arguments: @int ch@ = character to match |
| 765 | * @unsigned f@ = flags (@CTF_...@) to match |
| 766 | * |
| 767 | * Returns: The name of the character, or null if no match is found. |
| 768 | * |
| 769 | * Use: Looks up a name for a character. Specifically, it returns |
| 770 | * the first entry in the @chartab@ table which matches @ch@ and |
| 771 | * which has one of the flags @f@ set. |
| 772 | */ |
| 773 | |
| 774 | static const char *find_charname(int ch, unsigned f) |
| 775 | { |
| 776 | const struct chartab *ct; |
| 777 | |
| 778 | for (ct = chartab; ct->name; ct++) |
| 779 | if (ct->ch == ch && (ct->f&f)) return (ct->name); |
| 780 | return (0); |
| 781 | } |
| 782 | |
| 783 | /* --- @read_charname@ --- * |
| 784 | * |
| 785 | * Arguments: @int *ch_out@ = where to put the character |
| 786 | * @const char *p@ = character name |
| 787 | * @unsigned f@ = flags (@TCF_...@) |
| 788 | * |
| 789 | * Returns: Zero if a match was found, @-1@ if not. |
| 790 | * |
| 791 | * Use: Looks up a character by name. If @RCF_EOFOK@ is set in @f@, |
| 792 | * then the @EOF@ marker can be matched; otherwise it can't. |
| 793 | */ |
| 794 | |
| 795 | #define RCF_EOFOK 1u |
| 796 | static int read_charname(int *ch_out, const char *p, unsigned f) |
| 797 | { |
| 798 | const struct chartab *ct; |
| 799 | |
| 800 | for (ct = chartab; ct->name; ct++) |
| 801 | if (STRCMP(p, ==, ct->name) && ((f&RCF_EOFOK) || ct->ch >= 0)) |
| 802 | { *ch_out = ct->ch; return (0); } |
| 803 | return (-1); |
| 804 | } |
| 805 | |
| 806 | /* --- @format_charesc@ --- * |
| 807 | * |
| 808 | * Arguments: @const struct gprintf_ops *gops@ = print operations |
| 809 | * @void *go@ = print destination |
| 810 | * @int ch@ = character to format |
| 811 | * @unsigned f@ = flags (@FCF_...@) |
| 812 | * |
| 813 | * Returns: --- |
| 814 | * |
| 815 | * Use: Format a character as an escape sequence, possibly as part of |
| 816 | * a larger string. If @FCF_BRACE@ is set in @f@, then put |
| 817 | * braces around a `\x...' code, so that it's suitable for use |
| 818 | * in a longer string. |
| 819 | */ |
| 820 | |
| 821 | #define FCF_BRACE 1u |
| 822 | static void format_charesc(const struct gprintf_ops *gops, void *go, |
| 823 | int ch, unsigned f) |
| 824 | { |
| 825 | switch (ch) { |
| 826 | case '\a': gprintf(gops, go, "\\a"); break; |
| 827 | case '\b': gprintf(gops, go, "\\b"); break; |
| 828 | case '\x1b': gprintf(gops, go, "\\e"); break; |
| 829 | case '\f': gprintf(gops, go, "\\f"); break; |
| 830 | case '\r': gprintf(gops, go, "\\r"); break; |
| 831 | case '\n': gprintf(gops, go, "\\n"); break; |
| 832 | case '\t': gprintf(gops, go, "\\t"); break; |
| 833 | case '\v': gprintf(gops, go, "\\v"); break; |
| 834 | case '\\': gprintf(gops, go, "\\\\"); break; |
| 835 | case '\'': gprintf(gops, go, "\\'"); break; |
| 836 | case '\0': |
| 837 | if (f&FCF_BRACE) gprintf(gops, go, "\\{0}"); |
| 838 | else gprintf(gops, go, "\\0"); |
| 839 | break; |
| 840 | default: |
| 841 | if (f&FCF_BRACE) |
| 842 | gprintf(gops, go, "\\x{%0*x}", hex_width(UCHAR_MAX), ch); |
| 843 | else |
| 844 | gprintf(gops, go, "\\x%0*x", hex_width(UCHAR_MAX), ch); |
| 845 | break; |
| 846 | } |
| 847 | } |
| 848 | |
| 849 | /* --- @format_char@ --- * |
| 850 | * |
| 851 | * Arguments: @const struct gprintf_ops *gops@ = print operations |
| 852 | * @void *go@ = print destination |
| 853 | * @int ch@ = character to format |
| 854 | * |
| 855 | * Returns: --- |
| 856 | * |
| 857 | * Use: Format a single character. |
| 858 | */ |
| 859 | |
| 860 | static void format_char(const struct gprintf_ops *gops, void *go, int ch) |
| 861 | { |
| 862 | switch (ch) { |
| 863 | case '\\': case '\'': escape: |
| 864 | gprintf(gops, go, "'"); |
| 865 | format_charesc(gops, go, ch, 0); |
| 866 | gprintf(gops, go, "'"); |
| 867 | break; |
| 868 | default: |
| 869 | if (!isprint(ch)) goto escape; |
| 870 | gprintf(gops, go, "'%c'", ch); |
| 871 | break; |
| 872 | } |
| 873 | } |
| 874 | |
| 875 | /* --- @maybe_format_unsigned_char@, @maybe_format_signed_char@ --- * |
| 876 | * |
| 877 | * Arguments: @const struct gprintf_ops *gops@ = print operations |
| 878 | * @void *go@ = print destination |
| 879 | * @unsigned long u@ or @long i@ = an integer |
| 880 | * |
| 881 | * Returns: --- |
| 882 | * |
| 883 | * Use: Format a (signed or unsigned) integer as a character, if it's |
| 884 | * in range, printing something like `= 'q''. It's assumed that |
| 885 | * a comment marker has already been output. |
| 886 | */ |
| 887 | |
| 888 | static void maybe_format_unsigned_char |
| 889 | (const struct gprintf_ops *gops, void *go, unsigned long u) |
| 890 | { |
| 891 | const char *p; |
| 892 | |
| 893 | p = find_charname(u, CTF_PREFER); |
| 894 | if (p) gprintf(gops, go, " = %s", p); |
| 895 | if (u < UCHAR_MAX) |
| 896 | { gprintf(gops, go, " = "); format_char(gops, go, u); } |
| 897 | } |
| 898 | |
| 899 | static void maybe_format_signed_char |
| 900 | (const struct gprintf_ops *gops, void *go, long i) |
| 901 | { |
| 902 | const char *p; |
| 903 | |
| 904 | p = find_charname(i, CTF_PREFER); |
| 905 | if (p) gprintf(gops, go, " = %s", p); |
| 906 | if (0 <= i && i < UCHAR_MAX) |
| 907 | { gprintf(gops, go, " = "); format_char(gops, go, i); } |
| 908 | } |
| 909 | |
| 910 | /* --- @read_charesc@ --- * |
| 911 | * |
| 912 | * Arguments: @int *ch_out@ = where to put the result |
| 913 | * @struct tvec_state *tv@ = test vector state |
| 914 | * |
| 915 | * Returns: Zero on success, @-1@ on error. |
| 916 | * |
| 917 | * Use: Parse and convert an escape sequence from @tv@'s input |
| 918 | * stream, assuming that the initial `\' has already been read. |
| 919 | * Reports errors as appropriate. |
| 920 | */ |
| 921 | |
| 922 | static int read_charesc(int *ch_out, struct tvec_state *tv) |
| 923 | { |
| 924 | int ch, i, esc; |
| 925 | unsigned f = 0; |
| 926 | #define f_brace 1u |
| 927 | |
| 928 | ch = getc(tv->fp); |
| 929 | switch (ch) { |
| 930 | |
| 931 | /* Things we shouldn't find. */ |
| 932 | case EOF: case '\n': return (tvec_syntax(tv, ch, "string escape")); |
| 933 | |
| 934 | /* Single-character escapes. */ |
| 935 | case '\'': *ch_out = '\''; break; |
| 936 | case '\\': *ch_out = '\\'; break; |
| 937 | case '"': *ch_out = '"'; break; |
| 938 | case 'a': *ch_out = '\a'; break; |
| 939 | case 'b': *ch_out = '\b'; break; |
| 940 | case 'e': *ch_out = '\x1b'; break; |
| 941 | case 'f': *ch_out = '\f'; break; |
| 942 | case 'n': *ch_out = '\n'; break; |
| 943 | case 'r': *ch_out = '\r'; break; |
| 944 | case 't': *ch_out = '\t'; break; |
| 945 | case 'v': *ch_out = '\v'; break; |
| 946 | |
| 947 | /* Hex escapes, with and without braces. */ |
| 948 | case 'x': |
| 949 | ch = getc(tv->fp); |
| 950 | if (ch == '{') { f |= f_brace; ch = getc(tv->fp); } |
| 951 | else f &= ~f_brace; |
| 952 | esc = chtodig(ch); |
| 953 | if (esc < 0 || esc >= 16) return (tvec_syntax(tv, ch, "hex digit")); |
| 954 | for (;;) { |
| 955 | ch = getc(tv->fp); i = chtodig(ch); if (i < 0 || i >= 16) break; |
| 956 | esc = 16*esc + i; |
| 957 | if (esc > UCHAR_MAX) |
| 958 | return (tvec_error(tv, |
| 959 | "character code %d out of range", esc)); |
| 960 | } |
| 961 | if (!(f&f_brace)) ungetc(ch, tv->fp); |
| 962 | else if (ch != '}') return (tvec_syntax(tv, ch, "`}'")); |
| 963 | *ch_out = esc; |
| 964 | break; |
| 965 | |
| 966 | /* Other things, primarily octal escapes. */ |
| 967 | case '{': |
| 968 | f |= f_brace; ch = getc(tv->fp); |
| 969 | /* fall through */ |
| 970 | default: |
| 971 | if ('0' <= ch && ch < '8') { |
| 972 | i = 1; esc = ch - '0'; |
| 973 | for (;;) { |
| 974 | ch = getc(tv->fp); |
| 975 | if ('0' > ch || ch >= '8') { ungetc(ch, tv->fp); break; } |
| 976 | esc = 8*esc + ch - '0'; |
| 977 | i++; if (i >= 3) break; |
| 978 | } |
| 979 | if (f&f_brace) { |
| 980 | ch = getc(tv->fp); |
| 981 | if (ch != '}') return (tvec_syntax(tv, ch, "`}'")); |
| 982 | } |
| 983 | if (esc > UCHAR_MAX) |
| 984 | return (tvec_error(tv, |
| 985 | "character code %d out of range", esc)); |
| 986 | *ch_out = esc; break; |
| 987 | } else |
| 988 | return (tvec_syntax(tv, ch, "string escape")); |
| 989 | } |
| 990 | |
| 991 | /* Done. */ |
| 992 | return (0); |
| 993 | |
| 994 | #undef f_brace |
| 995 | } |
| 996 | |
| 997 | /* --- @read_quoted_string@ --- * |
| 998 | * |
| 999 | * Arguments: @dstr *d@ = string to write to |
| 1000 | * @int quote@ = initial quote, `'' or `"' |
| 1001 | * @struct tvec_state *tv@ = test vector state |
| 1002 | * |
| 1003 | * Returns: Zero on success, @-1@ on error. |
| 1004 | * |
| 1005 | * Use: Read the rest of a quoted string into @d@, reporting errors |
| 1006 | * as appropriate. |
| 1007 | * |
| 1008 | * A single-quoted string is entirely literal. A double-quoted |
| 1009 | * string may contain C-like escapes. |
| 1010 | */ |
| 1011 | |
| 1012 | static int read_quoted_string(dstr *d, int quote, struct tvec_state *tv) |
| 1013 | { |
| 1014 | int ch; |
| 1015 | |
| 1016 | for (;;) { |
| 1017 | ch = getc(tv->fp); |
| 1018 | switch (ch) { |
| 1019 | case EOF: case '\n': |
| 1020 | return (tvec_syntax(tv, ch, "`%c'", quote)); |
| 1021 | case '\\': |
| 1022 | if (quote == '\'') goto ordinary; |
| 1023 | ch = getc(tv->fp); if (ch == '\n') { tv->lno++; break; } |
| 1024 | ungetc(ch, tv->fp); if (read_charesc(&ch, tv)) return (-1); |
| 1025 | goto ordinary; |
| 1026 | default: |
| 1027 | if (ch == quote) goto end; |
| 1028 | ordinary: |
| 1029 | DPUTC(d, ch); |
| 1030 | break; |
| 1031 | } |
| 1032 | } |
| 1033 | |
| 1034 | end: |
| 1035 | DPUTZ(d); |
| 1036 | return (0); |
| 1037 | } |
| 1038 | |
| 1039 | /* --- @collect_bare@ --- * |
| 1040 | * |
| 1041 | * Arguments: @dstr *d@ = string to write to |
| 1042 | * @struct tvec_state *tv@ = test vector state |
| 1043 | * |
| 1044 | * Returns: Zero on success, @-1@ on error. |
| 1045 | * |
| 1046 | * Use: Read barewords and the whitespace between them. Stop when we |
| 1047 | * encounter something which can't start a bareword. |
| 1048 | */ |
| 1049 | |
| 1050 | static int collect_bare(dstr *d, struct tvec_state *tv) |
| 1051 | { |
| 1052 | size_t pos = d->len; |
| 1053 | enum { WORD, SPACE, ESCAPE }; unsigned s = WORD; |
| 1054 | int ch, rc; |
| 1055 | |
| 1056 | for (;;) { |
| 1057 | ch = getc(tv->fp); |
| 1058 | switch (ch) { |
| 1059 | case EOF: |
| 1060 | tvec_syntax(tv, ch, "bareword"); |
| 1061 | rc = -1; goto end; |
| 1062 | case '\n': |
| 1063 | if (s == ESCAPE) { tv->lno++; goto addch; } |
| 1064 | if (s == WORD) pos = d->len; |
| 1065 | ungetc(ch, tv->fp); if (tvec_nexttoken(tv)) { rc = -1; goto end; } |
| 1066 | DPUTC(d, ' '); s = SPACE; |
| 1067 | break; |
| 1068 | case '"': case '\'': case '!': case '#': case ')': case '}': case ']': |
| 1069 | if (s == SPACE) { ungetc(ch, tv->fp); goto done; } |
| 1070 | goto addch; |
| 1071 | case '\\': |
| 1072 | s = ESCAPE; |
| 1073 | break; |
| 1074 | default: |
| 1075 | if (s != ESCAPE && isspace(ch)) { |
| 1076 | if (s == WORD) pos = d->len; |
| 1077 | DPUTC(d, ch); s = SPACE; |
| 1078 | break; |
| 1079 | } |
| 1080 | addch: |
| 1081 | DPUTC(d, ch); s = WORD; |
| 1082 | } |
| 1083 | } |
| 1084 | |
| 1085 | done: |
| 1086 | if (s == SPACE) d->len = pos; |
| 1087 | DPUTZ(d); rc = 0; |
| 1088 | end: |
| 1089 | return (rc); |
| 1090 | } |
| 1091 | |
| 1092 | /* --- @set_up_encoding@ --- * |
| 1093 | * |
| 1094 | * Arguments: @const codec_class **ccl_out@ = where to put the class |
| 1095 | * @unsigned *f_out@ = where to put the flags |
| 1096 | * @unsigned code@ = the coding scheme to use (@TVEC_...@) |
| 1097 | * |
| 1098 | * Returns: --- |
| 1099 | * |
| 1100 | * Use: Helper for @read_compound_string@ below. |
| 1101 | * |
| 1102 | * Return the appropriate codec class and flags for @code@. |
| 1103 | * Leaves @*ccl_out@ null if the coding scheme doesn't have a |
| 1104 | * backing codec class (e.g., @TVCODE_BARE@). |
| 1105 | */ |
| 1106 | |
| 1107 | enum { TVCODE_BARE, TVCODE_HEX, TVCODE_BASE64, TVCODE_BASE32 }; |
| 1108 | static void set_up_encoding(const codec_class **ccl_out, unsigned *f_out, |
| 1109 | unsigned code) |
| 1110 | { |
| 1111 | switch (code) { |
| 1112 | case TVCODE_BARE: |
| 1113 | *ccl_out = 0; *f_out = 0; |
| 1114 | break; |
| 1115 | case TVCODE_HEX: |
| 1116 | *ccl_out = &hex_class; *f_out = CDCF_IGNCASE; |
| 1117 | break; |
| 1118 | case TVCODE_BASE32: |
| 1119 | *ccl_out = &base32_class; *f_out = CDCF_IGNCASE | CDCF_IGNEQPAD; |
| 1120 | break; |
| 1121 | case TVCODE_BASE64: |
| 1122 | *ccl_out = &base64_class; *f_out = CDCF_IGNEQPAD; |
| 1123 | break; |
| 1124 | default: |
| 1125 | abort(); |
| 1126 | } |
| 1127 | } |
| 1128 | |
| 1129 | /* --- @flush_codec@ --- * |
| 1130 | * |
| 1131 | * Arguments: @codec *cdc@ = a codec, or null |
| 1132 | * @dstr *d@ = output string |
| 1133 | * @struct tvec_state *tv@ = test vector state |
| 1134 | * |
| 1135 | * Returns: Zero on success, @-1@ on error. |
| 1136 | * |
| 1137 | * Use: Helper for @read_compound_string@ below. |
| 1138 | * |
| 1139 | * Flush out any final buffered material from @cdc@, and check |
| 1140 | * that it's in a good state. Frees the codec on success. Does |
| 1141 | * nothing if @cdc@ is null. |
| 1142 | */ |
| 1143 | |
| 1144 | static int flush_codec(codec *cdc, dstr *d, struct tvec_state *tv) |
| 1145 | { |
| 1146 | int err; |
| 1147 | |
| 1148 | if (cdc) { |
| 1149 | err = cdc->ops->code(cdc, 0, 0, d); |
| 1150 | if (err) |
| 1151 | return (tvec_error(tv, "invalid %s sequence end: %s", |
| 1152 | cdc->ops->c->name, codec_strerror(err))); |
| 1153 | cdc->ops->destroy(cdc); |
| 1154 | } |
| 1155 | return (0); |
| 1156 | } |
| 1157 | |
| 1158 | /* --- @read_compound_string@ --- * |
| 1159 | * |
| 1160 | * Arguments: @void **p_inout@ = address of output buffer pointer |
| 1161 | * @size_t *sz_inout@ = address of buffer size |
| 1162 | * @unsigned code@ = initial interpretation of barewords |
| 1163 | * @unsigned f@ = other flags (@RCSF_...@) |
| 1164 | * @struct tvec_state *tv@ = test vector state |
| 1165 | * |
| 1166 | * Returns: Zero on success, @-1@ on error. |
| 1167 | * |
| 1168 | * Use: Parse a compound string, i.e., a sequence of stringish pieces |
| 1169 | * which might be quoted strings, character names, or barewords |
| 1170 | * to be decoded accoding to @code@, interspersed with |
| 1171 | * additional directives. |
| 1172 | * |
| 1173 | * If the initial buffer pointer is non-null and sufficiently |
| 1174 | * large, then it will be reused; otherwise, it is freed and a |
| 1175 | * fresh, sufficiently large buffer is allocated and returned. |
| 1176 | */ |
| 1177 | |
| 1178 | #define RCSF_NESTED 1u |
| 1179 | static int read_compound_string(void **p_inout, size_t *sz_inout, |
| 1180 | unsigned code, unsigned f, |
| 1181 | struct tvec_state *tv) |
| 1182 | { |
| 1183 | const codec_class *ccl; unsigned cdf; |
| 1184 | codec *cdc; |
| 1185 | dstr d = DSTR_INIT, w = DSTR_INIT; |
| 1186 | char *p; |
| 1187 | const char *q; |
| 1188 | void *pp = 0; size_t sz; |
| 1189 | unsigned long n; |
| 1190 | int ch, err, rc; |
| 1191 | |
| 1192 | set_up_encoding(&ccl, &cdf, code); cdc = 0; |
| 1193 | |
| 1194 | if (tvec_nexttoken(tv)) return (tvec_syntax(tv, fgetc(tv->fp), "string")); |
| 1195 | do { |
| 1196 | ch = getc(tv->fp); |
| 1197 | switch (ch) { |
| 1198 | |
| 1199 | case ')': case ']': case '}': |
| 1200 | /* Close brackets. Leave these for recursive caller if there is one, |
| 1201 | * or just complain. |
| 1202 | */ |
| 1203 | |
| 1204 | if (!(f&RCSF_NESTED)) |
| 1205 | { rc = tvec_syntax(tv, ch, "string"); goto end; } |
| 1206 | ungetc(ch, tv->fp); goto done; |
| 1207 | |
| 1208 | case '"': case '\'': |
| 1209 | /* Quotes. Read a quoted string. */ |
| 1210 | |
| 1211 | if (cdc && flush_codec(cdc, &d, tv)) { rc = -1; goto end; } |
| 1212 | cdc = 0; |
| 1213 | if (read_quoted_string(&d, ch, tv)) { rc = -1; goto end; } |
| 1214 | break; |
| 1215 | |
| 1216 | case '#': |
| 1217 | /* A named character. */ |
| 1218 | |
| 1219 | ungetc(ch, tv->fp); |
| 1220 | if (cdc && flush_codec(cdc, &d, tv)) { rc = -1; goto end; } |
| 1221 | cdc = 0; |
| 1222 | DRESET(&w); tvec_readword(tv, &w, 0, ";", "character name"); |
| 1223 | if (read_charname(&ch, w.buf, RCF_EOFOK)) { |
| 1224 | rc = tvec_error(tv, "unknown character name `%s'", d.buf); |
| 1225 | goto end; |
| 1226 | } |
| 1227 | DPUTC(&d, ch); break; |
| 1228 | |
| 1229 | case '!': |
| 1230 | /* A magic keyword. */ |
| 1231 | |
| 1232 | if (cdc && flush_codec(cdc, &d, tv)) { rc = -1; goto end; } |
| 1233 | cdc = 0; |
| 1234 | ungetc(ch, tv->fp); |
| 1235 | DRESET(&w); tvec_readword(tv, &w, 0, ";", "`!'-keyword"); |
| 1236 | |
| 1237 | /* Change bareword coding system. */ |
| 1238 | if (STRCMP(w.buf, ==, "!bare")) |
| 1239 | { code = TVCODE_BARE; set_up_encoding(&ccl, &cdf, code); } |
| 1240 | else if (STRCMP(w.buf, ==, "!hex")) |
| 1241 | { code = TVCODE_HEX; set_up_encoding(&ccl, &cdf, code); } |
| 1242 | else if (STRCMP(w.buf, ==, "!base32")) |
| 1243 | { code = TVCODE_BASE32; set_up_encoding(&ccl, &cdf, code); } |
| 1244 | else if (STRCMP(w.buf, ==, "!base64")) |
| 1245 | { code = TVCODE_BASE64; set_up_encoding(&ccl, &cdf, code); } |
| 1246 | |
| 1247 | /* Repeated substrings. */ |
| 1248 | else if (STRCMP(w.buf, ==, "!repeat")) { |
| 1249 | if (tvec_nexttoken(tv)) { |
| 1250 | rc = tvec_syntax(tv, fgetc(tv->fp), "repeat count"); |
| 1251 | goto end; |
| 1252 | } |
| 1253 | DRESET(&w); |
| 1254 | if (tvec_readword(tv, &w, 0, ";{", "repeat count")) |
| 1255 | { rc = -1; goto end; } |
| 1256 | if (parse_unsigned_integer(&n, &q, w.buf)) { |
| 1257 | rc = tvec_error(tv, "invalid repeat count `%s'", w.buf); |
| 1258 | goto end; |
| 1259 | } |
| 1260 | if (*q) { rc = tvec_syntax(tv, *q, "`{'"); goto end; } |
| 1261 | if (tvec_nexttoken(tv)) |
| 1262 | { rc = tvec_syntax(tv, fgetc(tv->fp), "`{'"); goto end; } |
| 1263 | ch = getc(tv->fp); if (ch != '{') |
| 1264 | { rc = tvec_syntax(tv, ch, "`{'"); goto end; } |
| 1265 | sz = 0; |
| 1266 | if (read_compound_string(&pp, &sz, code, f | RCSF_NESTED, tv)) |
| 1267 | { rc = -1; goto end; } |
| 1268 | ch = getc(tv->fp); if (ch != '}') |
| 1269 | { rc = tvec_syntax(tv, ch, "`}'"); goto end; } |
| 1270 | if (sz) { |
| 1271 | if (n > (size_t)-1/sz) |
| 1272 | { rc = tvec_error(tv, "repeat size out of range"); goto end; } |
| 1273 | dstr_ensure(&d, n*sz); |
| 1274 | if (sz == 1) |
| 1275 | { memset(d.buf + d.len, *(unsigned char *)pp, n); d.len += n; } |
| 1276 | else |
| 1277 | for (; n--; d.len += sz) memcpy(d.buf + d.len, pp, sz); |
| 1278 | } |
| 1279 | xfree(pp); pp = 0; |
| 1280 | } |
| 1281 | |
| 1282 | /* Anything else is an error. */ |
| 1283 | else { |
| 1284 | tvec_error(tv, "unknown string keyword `%s'", w.buf); |
| 1285 | rc = -1; goto end; |
| 1286 | } |
| 1287 | break; |
| 1288 | |
| 1289 | default: |
| 1290 | /* A bareword. Process it according to the current coding system. */ |
| 1291 | |
| 1292 | switch (code) { |
| 1293 | case TVCODE_BARE: |
| 1294 | ungetc(ch, tv->fp); |
| 1295 | if (collect_bare(&d, tv)) goto done; |
| 1296 | break; |
| 1297 | default: |
| 1298 | assert(ccl); |
| 1299 | ungetc(ch, tv->fp); DRESET(&w); |
| 1300 | if (tvec_readword(tv, &w, 0, ";", |
| 1301 | "%s-encoded fragment", ccl->name)) |
| 1302 | { rc = -1; goto end; } |
| 1303 | if (!cdc) cdc = ccl->decoder(cdf); |
| 1304 | err = cdc->ops->code(cdc, w.buf, w.len, &d); |
| 1305 | if (err) { |
| 1306 | tvec_error(tv, "invalid %s fragment `%s': %s", |
| 1307 | ccl->name, w.buf, codec_strerror(err)); |
| 1308 | rc = -1; goto end; |
| 1309 | } |
| 1310 | break; |
| 1311 | } |
| 1312 | break; |
| 1313 | } |
| 1314 | } while (!tvec_nexttoken(tv)); |
| 1315 | |
| 1316 | done: |
| 1317 | /* Wrap things up. */ |
| 1318 | if (cdc && flush_codec(cdc, &d, tv)) { rc = -1; goto end; } |
| 1319 | cdc = 0; |
| 1320 | if (*sz_inout <= d.len) |
| 1321 | { xfree(*p_inout); *p_inout = xmalloc(d.len + 1); } |
| 1322 | p = *p_inout; memcpy(p, d.buf, d.len); p[d.len] = 0; *sz_inout = d.len; |
| 1323 | rc = 0; |
| 1324 | |
| 1325 | end: |
| 1326 | /* Clean up any debris. */ |
| 1327 | if (cdc) cdc->ops->destroy(cdc); |
| 1328 | if (pp) xfree(pp); |
| 1329 | dstr_destroy(&d); dstr_destroy(&w); |
| 1330 | return (rc); |
| 1331 | } |
| 1332 | |
| 1333 | /*----- Signed and unsigned integer types ---------------------------------*/ |
| 1334 | |
| 1335 | /* --- @init_int@, @init_uint@ --- * |
| 1336 | * |
| 1337 | * Arguments: @union tvec_regval *rv@ = register value |
| 1338 | * @const struct tvec_regdef *rd@ = register definition |
| 1339 | * |
| 1340 | * Returns: --- |
| 1341 | * |
| 1342 | * Use: Initialize a register value. |
| 1343 | * |
| 1344 | * Integer values are initialized to zero. |
| 1345 | */ |
| 1346 | |
| 1347 | static void init_int(union tvec_regval *rv, const struct tvec_regdef *rd) |
| 1348 | { rv->i = 0; } |
| 1349 | |
| 1350 | static void init_uint(union tvec_regval *rv, const struct tvec_regdef *rd) |
| 1351 | { rv->u = 0; } |
| 1352 | |
| 1353 | /* --- @eq_int@, @eq_uint@ --- * |
| 1354 | * |
| 1355 | * Arguments: @const union tvec_regval *rv0, *rv1@ = register values |
| 1356 | * @const struct tvec_regdef *rd@ = register definition |
| 1357 | * |
| 1358 | * Returns: Nonzero if the values are equal, zero if unequal |
| 1359 | * |
| 1360 | * Use: Compare register values for equality. |
| 1361 | */ |
| 1362 | |
| 1363 | static int eq_int(const union tvec_regval *rv0, const union tvec_regval *rv1, |
| 1364 | const struct tvec_regdef *rd) |
| 1365 | { return (rv0->i == rv1->i); } |
| 1366 | |
| 1367 | static int eq_uint(const union tvec_regval *rv0, |
| 1368 | const union tvec_regval *rv1, |
| 1369 | const struct tvec_regdef *rd) |
| 1370 | { return (rv0->u == rv1->u); } |
| 1371 | |
| 1372 | /* --- @tobuf_int@, @tobuf_uint@ --- * |
| 1373 | * |
| 1374 | * Arguments: @buf *b@ = buffer |
| 1375 | * @const union tvec_regval *rv@ = register value |
| 1376 | * @const struct tvec_regdef *rd@ = register definition |
| 1377 | * |
| 1378 | * Returns: Zero on success, %$-1$% on failure. |
| 1379 | * |
| 1380 | * Use: Serialize a register value to a buffer. |
| 1381 | * |
| 1382 | * Integer values are serialized as little-endian 64-bit signed |
| 1383 | * or unsigned integers. |
| 1384 | */ |
| 1385 | |
| 1386 | static int tobuf_int(buf *b, const union tvec_regval *rv, |
| 1387 | const struct tvec_regdef *rd) |
| 1388 | { return (signed_to_buf(b, rv->i)); } |
| 1389 | |
| 1390 | static int tobuf_uint(buf *b, const union tvec_regval *rv, |
| 1391 | const struct tvec_regdef *rd) |
| 1392 | { return (unsigned_to_buf(b, rv->u)); } |
| 1393 | |
| 1394 | /* --- @frombuf_int@, @frombuf_uint@ --- * |
| 1395 | * |
| 1396 | * Arguments: @buf *b@ = buffer |
| 1397 | * @union tvec_regval *rv@ = register value |
| 1398 | * @const struct tvec_regdef *rd@ = register definition |
| 1399 | * |
| 1400 | * Returns: Zero on success, %$-1$% on failure. |
| 1401 | * |
| 1402 | * Use: Deserialize a register value from a buffer. |
| 1403 | * |
| 1404 | * Integer values are serialized as 64-bit signed or unsigned |
| 1405 | * integers. |
| 1406 | */ |
| 1407 | |
| 1408 | static int frombuf_int(buf *b, union tvec_regval *rv, |
| 1409 | const struct tvec_regdef *rd) |
| 1410 | { return (signed_from_buf(b, &rv->i)); } |
| 1411 | |
| 1412 | static int frombuf_uint(buf *b, union tvec_regval *rv, |
| 1413 | const struct tvec_regdef *rd) |
| 1414 | { return (unsigned_from_buf(b, &rv->u)); } |
| 1415 | |
| 1416 | /* --- @parse_int@, @parse_uint@ --- * |
| 1417 | * |
| 1418 | * Arguments: @union tvec_regval *rv@ = register value |
| 1419 | * @const struct tvec_regdef *rd@ = register definition |
| 1420 | * @struct tvec_state *tv@ = test-vector state |
| 1421 | * |
| 1422 | * Returns: Zero on success, %$-1$% on error. |
| 1423 | * |
| 1424 | * Use: Parse a register value from an input file. |
| 1425 | * |
| 1426 | * Integers may be input in decimal, hex, binary, or octal, |
| 1427 | * following approximately usual conventions. |
| 1428 | * |
| 1429 | * * Signed integers may be preceded with a `+' or `-' sign. |
| 1430 | * |
| 1431 | * * Decimal integers are just a sequence of decimal digits |
| 1432 | * `0' ... `9'. |
| 1433 | * |
| 1434 | * * Octal integers are a sequence of digits `0' ... `7', |
| 1435 | * preceded by `0o' or `0O'. |
| 1436 | * |
| 1437 | * * Hexadecimal integers are a sequence of digits `0' |
| 1438 | * ... `9', `a' ... `f', or `A' ... `F', preceded by `0x' or |
| 1439 | * `0X'. |
| 1440 | * |
| 1441 | * * Radix-B integers are a sequence of digits `0' ... `9', |
| 1442 | * `a' ... `f', or `A' ... `F', each with value less than B, |
| 1443 | * preceded by `Br' or `BR', where 0 < B < 36 is expressed |
| 1444 | * in decimal without any leading `0' or internal |
| 1445 | * underscores `_'. |
| 1446 | * |
| 1447 | * * A digit sequence may contain internal underscore `_' |
| 1448 | * separators, but not before or after all of the digits; |
| 1449 | * and two consecutive `_' characters are not permitted. |
| 1450 | */ |
| 1451 | |
| 1452 | static int parse_int(union tvec_regval *rv, const struct tvec_regdef *rd, |
| 1453 | struct tvec_state *tv) |
| 1454 | { |
| 1455 | dstr d = DSTR_INIT; |
| 1456 | int rc; |
| 1457 | |
| 1458 | if (tvec_readword(tv, &d, 0, ";", "signed integer")) |
| 1459 | { rc = -1; goto end; } |
| 1460 | if (parse_signed(&rv->i, d.buf, rd->arg.p, tv)) { rc = -1; goto end; } |
| 1461 | if (tvec_flushtoeol(tv, 0)) { rc = -1; goto end; } |
| 1462 | rc = 0; |
| 1463 | end: |
| 1464 | dstr_destroy(&d); |
| 1465 | return (rc); |
| 1466 | } |
| 1467 | |
| 1468 | static int parse_uint(union tvec_regval *rv, const struct tvec_regdef *rd, |
| 1469 | struct tvec_state *tv) |
| 1470 | { |
| 1471 | dstr d = DSTR_INIT; |
| 1472 | int rc; |
| 1473 | |
| 1474 | if (tvec_readword(tv, &d, 0, ";", "unsigned integer")) |
| 1475 | { rc = -1; goto end; } |
| 1476 | if (parse_unsigned(&rv->u, d.buf, rd->arg.p, tv)) { rc = -1; goto end; } |
| 1477 | if (tvec_flushtoeol(tv, 0)) { rc = -1; goto end; } |
| 1478 | rc = 0; |
| 1479 | end: |
| 1480 | dstr_destroy(&d); |
| 1481 | return (rc); |
| 1482 | } |
| 1483 | |
| 1484 | /* --- @dump_int@, @dump_uint@ --- * |
| 1485 | * |
| 1486 | * Arguments: @const union tvec_regval *rv@ = register value |
| 1487 | * @const struct tvec_regdef *rd@ = register definition |
| 1488 | * @unsigned style@ = output style (@TVSF_...@) |
| 1489 | * @const struct gprintf_ops *gops@, @void *gp@ = format output |
| 1490 | * |
| 1491 | * Returns: --- |
| 1492 | * |
| 1493 | * Use: Dump a register value to the format output. |
| 1494 | * |
| 1495 | * Integer values are dumped in decimal and, unless compact |
| 1496 | * output is requested, hex, and maybe a character, as a |
| 1497 | * comment. |
| 1498 | */ |
| 1499 | |
| 1500 | static void dump_int(const union tvec_regval *rv, |
| 1501 | const struct tvec_regdef *rd, |
| 1502 | unsigned style, |
| 1503 | const struct gprintf_ops *gops, void *go) |
| 1504 | { |
| 1505 | |
| 1506 | gprintf(gops, go, "%ld", rv->i); |
| 1507 | if (!(style&TVSF_COMPACT)) { |
| 1508 | gprintf(gops, go, " ; = "); |
| 1509 | format_signed_hex(gops, go, rv->i); |
| 1510 | maybe_format_signed_char(gops, go, rv->i); |
| 1511 | } |
| 1512 | } |
| 1513 | |
| 1514 | static void dump_uint(const union tvec_regval *rv, |
| 1515 | const struct tvec_regdef *rd, |
| 1516 | unsigned style, |
| 1517 | const struct gprintf_ops *gops, void *go) |
| 1518 | { |
| 1519 | gprintf(gops, go, "%lu", rv->u); |
| 1520 | if (!(style&TVSF_COMPACT)) { |
| 1521 | gprintf(gops, go, " ; = "); |
| 1522 | format_unsigned_hex(gops, go, rv->u); |
| 1523 | maybe_format_unsigned_char(gops, go, rv->u); |
| 1524 | } |
| 1525 | } |
| 1526 | |
| 1527 | /* Integer type definitions. */ |
| 1528 | const struct tvec_regty tvty_int = { |
| 1529 | init_int, trivial_release, eq_int, |
| 1530 | tobuf_int, frombuf_int, |
| 1531 | parse_int, dump_int |
| 1532 | }; |
| 1533 | const struct tvec_regty tvty_uint = { |
| 1534 | init_uint, trivial_release, eq_uint, |
| 1535 | tobuf_uint, frombuf_uint, |
| 1536 | parse_uint, dump_uint |
| 1537 | }; |
| 1538 | |
| 1539 | /* Predefined integer ranges. */ |
| 1540 | const struct tvec_irange |
| 1541 | tvrange_schar = { SCHAR_MIN, SCHAR_MAX }, |
| 1542 | tvrange_short = { SHRT_MIN, SHRT_MAX }, |
| 1543 | tvrange_int = { INT_MIN, INT_MAX }, |
| 1544 | tvrange_long = { LONG_MIN, LONG_MAX }, |
| 1545 | tvrange_sbyte = { -128, 127 }, |
| 1546 | tvrange_i16 = { -32768, +32767 }, |
| 1547 | tvrange_i32 = { -2147483648, 2147483647 }; |
| 1548 | const struct tvec_urange |
| 1549 | tvrange_uchar = { 0, UCHAR_MAX }, |
| 1550 | tvrange_ushort = { 0, USHRT_MAX }, |
| 1551 | tvrange_uint = { 0, UINT_MAX }, |
| 1552 | tvrange_ulong = { 0, ULONG_MAX }, |
| 1553 | tvrange_size = { 0, (size_t)-1 }, |
| 1554 | tvrange_byte = { 0, 255 }, |
| 1555 | tvrange_u16 = { 0, 65535 }, |
| 1556 | tvrange_u32 = { 0, 4294967296 }; |
| 1557 | |
| 1558 | /* --- @tvec_claimeq_int@ --- * |
| 1559 | * |
| 1560 | * Arguments: @struct tvec_state *tv@ = test-vector state |
| 1561 | * @long i0, i1@ = two signed integers |
| 1562 | * @const char *file@, @unsigned @lno@ = calling file and line |
| 1563 | * @const char *expr@ = the expression to quote on failure |
| 1564 | * |
| 1565 | * Returns: Nonzero if @i0@ and @i1@ are equal, otherwise zero. |
| 1566 | * |
| 1567 | * Use: Check that values of @i0@ and @i1@ are equal. As for |
| 1568 | * @tvec_claim@ above, a test case is automatically begun and |
| 1569 | * ended if none is already underway. If the values are |
| 1570 | * unequal, then @tvec_fail@ is called, quoting @expr@, and the |
| 1571 | * mismatched values are dumped: @i0@ is printed as the output |
| 1572 | * value and @i1@ is printed as the input reference. |
| 1573 | */ |
| 1574 | |
| 1575 | int tvec_claimeq_int(struct tvec_state *tv, long i0, long i1, |
| 1576 | const char *file, unsigned lno, const char *expr) |
| 1577 | { |
| 1578 | tv->out[0].v.i = i0; tv->in[0].v.i = i1; |
| 1579 | return (tvec_claimeq(tv, &tvty_int, 0, file, lno, expr)); |
| 1580 | } |
| 1581 | |
| 1582 | /* --- @tvec_claimeq_uint@ --- * |
| 1583 | * |
| 1584 | * Arguments: @struct tvec_state *tv@ = test-vector state |
| 1585 | * @unsigned long u0, u1@ = two unsigned integers |
| 1586 | * @const char *file@, @unsigned @lno@ = calling file and line |
| 1587 | * @const char *expr@ = the expression to quote on failure |
| 1588 | * |
| 1589 | * Returns: Nonzero if @u0@ and @u1@ are equal, otherwise zero. |
| 1590 | * |
| 1591 | * Use: Check that values of @u0@ and @u1@ are equal. As for |
| 1592 | * @tvec_claim@ above, a test case is automatically begun and |
| 1593 | * ended if none is already underway. If the values are |
| 1594 | * unequal, then @tvec_fail@ is called, quoting @expr@, and the |
| 1595 | * mismatched values are dumped: @u0@ is printed as the output |
| 1596 | * value and @u1@ is printed as the input reference. |
| 1597 | */ |
| 1598 | |
| 1599 | int tvec_claimeq_uint(struct tvec_state *tv, |
| 1600 | unsigned long u0, unsigned long u1, |
| 1601 | const char *file, unsigned lno, const char *expr) |
| 1602 | { |
| 1603 | tv->out[0].v.u = u0; tv->in[0].v.u = u1; |
| 1604 | return (tvec_claimeq(tv, &tvty_uint, 0, file, lno, expr)); |
| 1605 | } |
| 1606 | |
| 1607 | /*----- Floating-point type -----------------------------------------------*/ |
| 1608 | |
| 1609 | /* --- @int_float@ --- * |
| 1610 | * |
| 1611 | * Arguments: @union tvec_regval *rv@ = register value |
| 1612 | * @const struct tvec_regdef *rd@ = register definition |
| 1613 | * |
| 1614 | * Returns: --- |
| 1615 | * |
| 1616 | * Use: Initialize a register value. |
| 1617 | * |
| 1618 | * Floating-point values are initialized to zero. |
| 1619 | */ |
| 1620 | |
| 1621 | static void init_float(union tvec_regval *rv, const struct tvec_regdef *rd) |
| 1622 | { rv->f = 0.0; } |
| 1623 | |
| 1624 | /* --- @eq_float@ --- * |
| 1625 | * |
| 1626 | * Arguments: @const union tvec_regval *rv0, *rv1@ = register values |
| 1627 | * @const struct tvec_regdef *rd@ = register definition |
| 1628 | * |
| 1629 | * Returns: Nonzero if the values are equal, zero if unequal |
| 1630 | * |
| 1631 | * Use: Compare register values for equality. |
| 1632 | * |
| 1633 | * Floating-point values may be considered equal if their |
| 1634 | * absolute or relative difference is sufficiently small, as |
| 1635 | * described in the register definition. |
| 1636 | */ |
| 1637 | |
| 1638 | static int eq_float(const union tvec_regval *rv0, |
| 1639 | const union tvec_regval *rv1, |
| 1640 | const struct tvec_regdef *rd) |
| 1641 | { return (eqish_floating_p(rv0->f, rv1->f, rd->arg.p)); } |
| 1642 | |
| 1643 | /* --- @tobuf_float@ --- * |
| 1644 | * |
| 1645 | * Arguments: @buf *b@ = buffer |
| 1646 | * @const union tvec_regval *rv@ = register value |
| 1647 | * @const struct tvec_regdef *rd@ = register definition |
| 1648 | * |
| 1649 | * Returns: Zero on success, %$-1$% on failure. |
| 1650 | * |
| 1651 | * Use: Serialize a register value to a buffer. |
| 1652 | * |
| 1653 | * Floating-point values are serialized as little-endian |
| 1654 | * IEEE 754 Binary64. |
| 1655 | */ |
| 1656 | |
| 1657 | static int tobuf_float(buf *b, const union tvec_regval *rv, |
| 1658 | const struct tvec_regdef *rd) |
| 1659 | { return (buf_putf64l(b, rv->f)); } |
| 1660 | |
| 1661 | /* --- @frombuf_float@ --- * |
| 1662 | * |
| 1663 | * Arguments: @buf *b@ = buffer |
| 1664 | * @union tvec_regval *rv@ = register value |
| 1665 | * @const struct tvec_regdef *rd@ = register definition |
| 1666 | * |
| 1667 | * Returns: Zero on success, %$-1$% on failure. |
| 1668 | * |
| 1669 | * Use: Deserialize a register value from a buffer. |
| 1670 | * |
| 1671 | * Floating-point values are serialized as little-endian |
| 1672 | * IEEE 754 Binary64. |
| 1673 | */ |
| 1674 | |
| 1675 | static int frombuf_float(buf *b, union tvec_regval *rv, |
| 1676 | const struct tvec_regdef *rd) |
| 1677 | { return (buf_getf64l(b, &rv->f)); } |
| 1678 | |
| 1679 | /* --- @parse_float@ --- * |
| 1680 | * |
| 1681 | * Arguments: @union tvec_regval *rv@ = register value |
| 1682 | * @const struct tvec_regdef *rd@ = register definition |
| 1683 | * @struct tvec_state *tv@ = test-vector state |
| 1684 | * |
| 1685 | * Returns: Zero on success, %$-1$% on error. |
| 1686 | * |
| 1687 | * Use: Parse a register value from an input file. |
| 1688 | * |
| 1689 | * Floating-point values are either NaN (%|#nan|%, if supported |
| 1690 | * by the platform); positive or negative infinity (%|#inf|%, |
| 1691 | * %|+#inf|%, or %|#+inf|% (preferring the last), and %|-#inf|% |
| 1692 | * or %|#-inf|% (preferring the latter), if supported by the |
| 1693 | * platform); or a number in strtod(3) syntax. |
| 1694 | */ |
| 1695 | |
| 1696 | static int parse_float(union tvec_regval *rv, const struct tvec_regdef *rd, |
| 1697 | struct tvec_state *tv) |
| 1698 | { |
| 1699 | dstr d = DSTR_INIT; |
| 1700 | int rc; |
| 1701 | |
| 1702 | if (tvec_readword(tv, &d, 0, ";", "floating-point number")) |
| 1703 | { rc = -1; goto end; } |
| 1704 | if (parse_floating(&rv->f, 0, d.buf, rd->arg.p, tv)) |
| 1705 | { rc = -1; goto end; } |
| 1706 | if (tvec_flushtoeol(tv, 0)) { rc = -1; goto end; } |
| 1707 | rc = 0; |
| 1708 | end: |
| 1709 | dstr_destroy(&d); |
| 1710 | return (rc); |
| 1711 | } |
| 1712 | |
| 1713 | /* --- @dump_float@ --- * |
| 1714 | * |
| 1715 | * Arguments: @const union tvec_regval *rv@ = register value |
| 1716 | * @const struct tvec_regdef *rd@ = register definition |
| 1717 | * @unsigned style@ = output style (@TVSF_...@) |
| 1718 | * @const struct gprintf_ops *gops@, @void *gp@ = format output |
| 1719 | * |
| 1720 | * Returns: --- |
| 1721 | * |
| 1722 | * Use: Dump a register value to the format output. |
| 1723 | * |
| 1724 | * Floating-point values are dumped in decimal or as a special |
| 1725 | * token beginning with `%|#|%'. Some effort is taken to ensure |
| 1726 | * that the output is sufficient to uniquely identify the |
| 1727 | * original value, but, honestly, C makes this really hard. |
| 1728 | */ |
| 1729 | |
| 1730 | static void dump_float(const union tvec_regval *rv, |
| 1731 | const struct tvec_regdef *rd, |
| 1732 | unsigned style, |
| 1733 | const struct gprintf_ops *gops, void *go) |
| 1734 | { format_floating(gops, go, rv->f); } |
| 1735 | |
| 1736 | /* Floating-point type definition. */ |
| 1737 | const struct tvec_regty tvty_float = { |
| 1738 | init_float, trivial_release, eq_float, |
| 1739 | tobuf_float, frombuf_float, |
| 1740 | parse_float, dump_float |
| 1741 | }; |
| 1742 | |
| 1743 | /* Predefined floating-point ranges. */ |
| 1744 | const struct tvec_floatinfo |
| 1745 | tvflt_finite = { TVFF_EXACT, -DBL_MAX, DBL_MAX, 0.0 }, |
| 1746 | tvflt_nonneg = { TVFF_EXACT, 0, DBL_MAX, 0.0 }; |
| 1747 | |
| 1748 | /* --- @tvec_claimeqish_float@ --- * |
| 1749 | * |
| 1750 | * Arguments: @struct tvec_state *tv@ = test-vector state |
| 1751 | * @double f0, f1@ = two floating-point numbers |
| 1752 | * @unsigned f@ = flags (@TVFF_...@) |
| 1753 | * @double delta@ = maximum tolerable difference |
| 1754 | * @const char *file@, @unsigned @lno@ = calling file and line |
| 1755 | * @const char *expr@ = the expression to quote on failure |
| 1756 | * |
| 1757 | * Returns: Nonzero if @f0@ and @u1@ are sufficiently close, otherwise |
| 1758 | * zero. |
| 1759 | * |
| 1760 | * Use: Check that values of @f0@ and @f1@ are sufficiently close. |
| 1761 | * As for @tvec_claim@ above, a test case is automatically begun |
| 1762 | * and ended if none is already underway. If the values are |
| 1763 | * too far apart, then @tvec_fail@ is called, quoting @expr@, |
| 1764 | * and the mismatched values are dumped: @f0@ is printed as the |
| 1765 | * output value and @f1@ is printed as the input reference. |
| 1766 | * |
| 1767 | * The details for the comparison are as follows. |
| 1768 | * |
| 1769 | * * A NaN value matches any other NaN, and nothing else. |
| 1770 | * |
| 1771 | * * An infinity matches another infinity of the same sign, |
| 1772 | * and nothing else. |
| 1773 | * |
| 1774 | * * If @f&TVFF_EQMASK@ is @TVFF_EXACT@, then any |
| 1775 | * representable number matches only itself: in particular, |
| 1776 | * positive and negative zero are considered distinct. |
| 1777 | * (This allows tests to check that they land on the correct |
| 1778 | * side of branch cuts, for example.) |
| 1779 | * |
| 1780 | * * If @f&TVFF_EQMASK@ is @TVFF_ABSDELTA@, then %$x$% matches |
| 1781 | * %$y$% when %$|x - y| < \delta$%. |
| 1782 | * |
| 1783 | * * If @f&TVFF_EQMASK@ is @TVFF_RELDELTA@, then %$x$% matches |
| 1784 | * %$y$% when %$|1 - y/x| < \delta$%. (Note that this |
| 1785 | * criterion is asymmetric FIXME |
| 1786 | */ |
| 1787 | |
| 1788 | int tvec_claimeqish_float(struct tvec_state *tv, |
| 1789 | double f0, double f1, unsigned f, double delta, |
| 1790 | const char *file, unsigned lno, |
| 1791 | const char *expr) |
| 1792 | { |
| 1793 | struct tvec_floatinfo fi; |
| 1794 | union tvec_misc arg; |
| 1795 | |
| 1796 | fi.f = f; fi.min = fi.max = 0.0; fi.delta = delta; arg.p = &fi; |
| 1797 | tv->out[0].v.f = f0; tv->in[0].v.f = f1; |
| 1798 | return (tvec_claimeq(tv, &tvty_float, &arg, file, lno, expr)); |
| 1799 | } |
| 1800 | |
| 1801 | /* --- @tvec_claimeq_float@ --- * |
| 1802 | * |
| 1803 | * Arguments: @struct tvec_state *tv@ = test-vector state |
| 1804 | * @double f0, f1@ = two floating-point numbers |
| 1805 | * @const char *file@, @unsigned @lno@ = calling file and line |
| 1806 | * @const char *expr@ = the expression to quote on failure |
| 1807 | * |
| 1808 | * Returns: Nonzero if @f0@ and @u1@ are identical, otherwise zero. |
| 1809 | * |
| 1810 | * Use: Check that values of @f0@ and @f1@ are identical. The |
| 1811 | * function is exactly equivalent to @tvec_claimeqish_float@ |
| 1812 | * with @f == TVFF_EXACT@. |
| 1813 | */ |
| 1814 | |
| 1815 | int tvec_claimeq_float(struct tvec_state *tv, |
| 1816 | double f0, double f1, |
| 1817 | const char *file, unsigned lno, |
| 1818 | const char *expr) |
| 1819 | { |
| 1820 | return (tvec_claimeqish_float(tv, f0, f1, TVFF_EXACT, 0.0, |
| 1821 | file, lno, expr)); |
| 1822 | } |
| 1823 | |
| 1824 | /*----- Durations ---------------------------------------------------------*/ |
| 1825 | |
| 1826 | /* A duration is a floating-point number of seconds. Initialization and |
| 1827 | * teardown, equality comparison, and serialization are as for floating-point |
| 1828 | * values. |
| 1829 | */ |
| 1830 | |
| 1831 | static const struct duration_unit { |
| 1832 | const char *unit; |
| 1833 | double scale; |
| 1834 | unsigned f; |
| 1835 | #define DUF_PREFER 1u |
| 1836 | } duration_units[] = { |
| 1837 | { "Ys", 1e+24, 0 }, |
| 1838 | { "Zs", 1e+21, 0 }, |
| 1839 | { "Es", 1e+18, 0 }, |
| 1840 | { "Ps", 1e+15, 0 }, |
| 1841 | { "Ts", 1e+12, 0 }, |
| 1842 | { "Gs", 1e+9, 0 }, |
| 1843 | { "Ms", 1e+6, 0 }, |
| 1844 | { "ks", 1e+3, 0 }, |
| 1845 | { "hs", 1e+2, 0 }, |
| 1846 | { "das", 1e+1, 0 }, |
| 1847 | |
| 1848 | { "yr", 31557600.0, DUF_PREFER }, |
| 1849 | { "y", 31557600.0, 0 }, |
| 1850 | { "day", 86400.0, DUF_PREFER }, |
| 1851 | { "dy", 86400.0, 0 }, |
| 1852 | { "d", 86400.0, 0 }, |
| 1853 | { "hr", 3600.0, DUF_PREFER }, |
| 1854 | { "hour", 3600.0, 0 }, |
| 1855 | { "h", 3600.0, 0 }, |
| 1856 | { "min", 60.0, DUF_PREFER }, |
| 1857 | { "m", 60.0, 0 }, |
| 1858 | |
| 1859 | { "s", 1.0, DUF_PREFER }, |
| 1860 | { "sec", 1.0, 0 }, |
| 1861 | |
| 1862 | { "ds", 1e-1, 0 }, |
| 1863 | { "cs", 1e-2, 0 }, |
| 1864 | { "ms", 1e-3, DUF_PREFER }, |
| 1865 | { "µs", 1e-6, DUF_PREFER }, |
| 1866 | { "ns", 1e-9, DUF_PREFER }, |
| 1867 | { "ps", 1e-12, DUF_PREFER }, |
| 1868 | { "fs", 1e-15, DUF_PREFER }, |
| 1869 | { "as", 1e-18, DUF_PREFER }, |
| 1870 | { "zs", 1e-21, DUF_PREFER }, |
| 1871 | { "ys", 1e-24, DUF_PREFER }, |
| 1872 | |
| 1873 | { 0 } |
| 1874 | }; |
| 1875 | |
| 1876 | /* --- @parse_duration@ --- * |
| 1877 | * |
| 1878 | * Arguments: @union tvec_regval *rv@ = register value |
| 1879 | * @const struct tvec_regdef *rd@ = register definition |
| 1880 | * @struct tvec_state *tv@ = test-vector state |
| 1881 | * |
| 1882 | * Returns: Zero on success, %$-1$% on error. |
| 1883 | * |
| 1884 | * Use: Parse a register value from an input file. |
| 1885 | * |
| 1886 | * Duration values are finite nonnegative floating-point |
| 1887 | * numbers in @strtod@ syntax, optionally followed by a unit . |
| 1888 | */ |
| 1889 | |
| 1890 | static int parse_duration(union tvec_regval *rv, |
| 1891 | const struct tvec_regdef *rd, |
| 1892 | struct tvec_state *tv) |
| 1893 | { |
| 1894 | const struct duration_unit *u; |
| 1895 | const char *q; |
| 1896 | dstr d = DSTR_INIT; |
| 1897 | double t; |
| 1898 | int rc; |
| 1899 | |
| 1900 | if (tvec_readword(tv, &d, 0, ";", "duration")) { rc = -1; goto end; } |
| 1901 | if (parse_floating(&t, &q, d.buf, |
| 1902 | rd->arg.p ? rd->arg.p : &tvflt_nonneg, tv)) |
| 1903 | { rc = -1; goto end; } |
| 1904 | |
| 1905 | if (!*q) tvec_readword(tv, &d, &q, ";", 0); |
| 1906 | if (*q) { |
| 1907 | for (u = duration_units; u->unit; u++) |
| 1908 | if (STRCMP(q, ==, u->unit)) { t *= u->scale; goto found_unit; } |
| 1909 | rc = tvec_syntax(tv, *q, "end-of-line"); goto end; |
| 1910 | found_unit:; |
| 1911 | } |
| 1912 | |
| 1913 | if (tvec_flushtoeol(tv, 0)) { rc = -1; goto end; } |
| 1914 | rv->f = t; rc = 0; |
| 1915 | end: |
| 1916 | dstr_destroy(&d); |
| 1917 | return (rc); |
| 1918 | } |
| 1919 | |
| 1920 | /* --- @dump_duration@ --- * |
| 1921 | * |
| 1922 | * Arguments: @const union tvec_regval *rv@ = register value |
| 1923 | * @const struct tvec_regdef *rd@ = register definition |
| 1924 | * @unsigned style@ = output style (@TVSF_...@) |
| 1925 | * @const struct gprintf_ops *gops@, @void *gp@ = format output |
| 1926 | * |
| 1927 | * Returns: --- |
| 1928 | * |
| 1929 | * Use: Dump a register value to the format output. |
| 1930 | * |
| 1931 | * Durations are dumped as a human-palatable scaled value with |
| 1932 | * unit, and, if compact style is not requested, as a raw number |
| 1933 | * of seconds at full precision as a comment. |
| 1934 | */ |
| 1935 | |
| 1936 | static void dump_duration(const union tvec_regval *rv, |
| 1937 | const struct tvec_regdef *rd, |
| 1938 | unsigned style, |
| 1939 | const struct gprintf_ops *gops, void *go) |
| 1940 | { |
| 1941 | const struct duration_unit *u; |
| 1942 | double t = rv->f; |
| 1943 | |
| 1944 | if (!t) u = 0; |
| 1945 | else { |
| 1946 | for (u = duration_units; u->scale > t && u[1].unit; u++); |
| 1947 | t /= u->scale; |
| 1948 | } |
| 1949 | |
| 1950 | gprintf(gops, go, "%.4g %s", t, u ? u->unit : "s"); |
| 1951 | if (!(style&TVSF_COMPACT)) { |
| 1952 | gprintf(gops, go, "; = "); |
| 1953 | format_floating(gops, go, rv->f); |
| 1954 | gprintf(gops, go, " s"); |
| 1955 | } |
| 1956 | } |
| 1957 | |
| 1958 | /* Duration type definition. */ |
| 1959 | const struct tvec_regty tvty_duration = { |
| 1960 | init_float, trivial_release, eq_float, |
| 1961 | tobuf_float, frombuf_float, |
| 1962 | parse_duration, dump_duration |
| 1963 | }; |
| 1964 | |
| 1965 | /*----- Enumerations ------------------------------------------------------*/ |
| 1966 | |
| 1967 | /* --- @init_tenum@ --- * |
| 1968 | * |
| 1969 | * Arguments: @union tvec_regval *rv@ = register value |
| 1970 | * @const struct tvec_regdef *rd@ = register definition |
| 1971 | * |
| 1972 | * Returns: --- |
| 1973 | * |
| 1974 | * Use: Initialize a register value. |
| 1975 | * |
| 1976 | * Integer and floating-point enumeration values are initialized |
| 1977 | * as their underlying representations. Pointer enumerations |
| 1978 | * are initialized to %|#nil|%. |
| 1979 | */ |
| 1980 | |
| 1981 | #define init_ienum init_int |
| 1982 | #define init_uenum init_uint |
| 1983 | #define init_fenum init_float |
| 1984 | |
| 1985 | static void init_penum(union tvec_regval *rv, const struct tvec_regdef *rd) |
| 1986 | { rv->p = 0; } |
| 1987 | |
| 1988 | /* --- @eq_tenum@ --- * |
| 1989 | * |
| 1990 | * Arguments: @const union tvec_regval *rv0, *rv1@ = register values |
| 1991 | * @const struct tvec_regdef *rd@ = register definition |
| 1992 | * |
| 1993 | * Returns: Nonzero if the values are equal, zero if unequal |
| 1994 | * |
| 1995 | * Use: Compare register values for equality. |
| 1996 | * |
| 1997 | * Integer and floating-point enumeration values are compared as |
| 1998 | * their underlying representations; in particular, floating- |
| 1999 | * point enumerations may compare equal if their absolute or |
| 2000 | * relative difference is sufficiently small. Pointer |
| 2001 | * enumerations are compared as pointers. |
| 2002 | */ |
| 2003 | |
| 2004 | #define eq_ienum eq_int |
| 2005 | #define eq_uenum eq_uint |
| 2006 | |
| 2007 | static int eq_fenum(const union tvec_regval *rv0, |
| 2008 | const union tvec_regval *rv1, |
| 2009 | const struct tvec_regdef *rd) |
| 2010 | { |
| 2011 | const struct tvec_fenuminfo *ei = rd->arg.p; |
| 2012 | return (eqish_floating_p(rv0->f, rv1->f, ei->fi)); |
| 2013 | } |
| 2014 | |
| 2015 | static int eq_penum(const union tvec_regval *rv0, |
| 2016 | const union tvec_regval *rv1, |
| 2017 | const struct tvec_regdef *rd) |
| 2018 | { return (rv0->p == rv1->p); } |
| 2019 | |
| 2020 | /* --- @tobuf_tenum@ --- * |
| 2021 | * |
| 2022 | * Arguments: @buf *b@ = buffer |
| 2023 | * @const union tvec_regval *rv@ = register value |
| 2024 | * @const struct tvec_regdef *rd@ = register definition |
| 2025 | * |
| 2026 | * Returns: Zero on success, %$-1$% on failure. |
| 2027 | * |
| 2028 | * Use: Serialize a register value to a buffer. |
| 2029 | * |
| 2030 | * Integer and floating-point enumeration values are serialized |
| 2031 | * as their underlying representations. Pointer enumerations |
| 2032 | * are serialized as the signed integer index into the |
| 2033 | * association table; %|#nil|% serializes as %$-1$%, and |
| 2034 | * unrecognized pointers cause failure. |
| 2035 | */ |
| 2036 | |
| 2037 | #define tobuf_ienum tobuf_int |
| 2038 | #define tobuf_uenum tobuf_uint |
| 2039 | #define tobuf_fenum tobuf_float |
| 2040 | |
| 2041 | static int tobuf_penum(buf *b, const union tvec_regval *rv, |
| 2042 | const struct tvec_regdef *rd) |
| 2043 | { |
| 2044 | const struct tvec_penuminfo *pei = rd->arg.p; |
| 2045 | const struct tvec_passoc *pa; |
| 2046 | long i; |
| 2047 | |
| 2048 | for (pa = pei->av, i = 0; pa->tag; pa++, i++) |
| 2049 | if (pa->p == rv->p) goto found; |
| 2050 | if (!rv->p) i = -1; |
| 2051 | else return (-1); |
| 2052 | found: |
| 2053 | return (signed_to_buf(b, i)); |
| 2054 | } |
| 2055 | |
| 2056 | /* --- @frombuf_tenum@ --- * |
| 2057 | * |
| 2058 | * Arguments: @buf *b@ = buffer |
| 2059 | * @union tvec_regval *rv@ = register value |
| 2060 | * @const struct tvec_regdef *rd@ = register definition |
| 2061 | * |
| 2062 | * Returns: Zero on success, %$-1$% on failure. |
| 2063 | * |
| 2064 | * Use: Deserialize a register value from a buffer. |
| 2065 | * |
| 2066 | * Integer and floating-point enumeration values are serialized |
| 2067 | * as their underlying representations. Pointer enumerations |
| 2068 | * are serialized as the signed integer index into the |
| 2069 | * association table; %|#nil|% serializes as %$-1$%; out-of- |
| 2070 | * range indices cause failure. |
| 2071 | */ |
| 2072 | |
| 2073 | #define frombuf_ienum frombuf_int |
| 2074 | #define frombuf_uenum frombuf_uint |
| 2075 | #define frombuf_fenum frombuf_float |
| 2076 | static int frombuf_penum(buf *b, union tvec_regval *rv, |
| 2077 | const struct tvec_regdef *rd) |
| 2078 | { |
| 2079 | const struct tvec_penuminfo *pei = rd->arg.p; |
| 2080 | const struct tvec_passoc *pa; |
| 2081 | long i, n; |
| 2082 | |
| 2083 | for (pa = pei->av, n = 0; pa->tag; pa++, n++); |
| 2084 | if (signed_from_buf(b, &i)) return (-1); |
| 2085 | if (0 <= i && i < n) rv->p = (/*unconst*/ void *)pei->av[i].p; |
| 2086 | else if (i == -1) rv->p = 0; |
| 2087 | else { buf_break(b); return (-1); } |
| 2088 | return (0); |
| 2089 | } |
| 2090 | |
| 2091 | /* --- @parse_tenum@ --- * |
| 2092 | * |
| 2093 | * Arguments: @union tvec_regval *rv@ = register value |
| 2094 | * @const struct tvec_regdef *rd@ = register definition |
| 2095 | * @struct tvec_state *tv@ = test-vector state |
| 2096 | * |
| 2097 | * Returns: Zero on success, %$-1$% on error. |
| 2098 | * |
| 2099 | * Use: Parse a register value from an input file. |
| 2100 | * |
| 2101 | * An enumerated value may be given by name or as a literal |
| 2102 | * value. For enumerations based on numeric types, the literal |
| 2103 | * values can be written in the same syntax as the underlying |
| 2104 | * values. For enumerations based on pointers, the only |
| 2105 | * permitted literal is %|#nil|%, which denotes a null pointer. |
| 2106 | */ |
| 2107 | |
| 2108 | #define DEFPARSE_ENUM(tag_, ty, slot) \ |
| 2109 | static int parse_##slot##enum(union tvec_regval *rv, \ |
| 2110 | const struct tvec_regdef *rd, \ |
| 2111 | struct tvec_state *tv) \ |
| 2112 | { \ |
| 2113 | const struct tvec_##slot##enuminfo *ei = rd->arg.p; \ |
| 2114 | const struct tvec_##slot##assoc *a; \ |
| 2115 | dstr d = DSTR_INIT; \ |
| 2116 | int rc; \ |
| 2117 | \ |
| 2118 | if (tvec_readword(tv, &d, 0, ";", "enumeration tag or " LITSTR_##tag_)) \ |
| 2119 | { rc = -1; goto end; } \ |
| 2120 | for (a = ei->av; a->tag; a++) \ |
| 2121 | if (STRCMP(a->tag, ==, d.buf)) { FOUND_##tag_ goto done; } \ |
| 2122 | MISSING_##tag_ \ |
| 2123 | done: \ |
| 2124 | if (tvec_flushtoeol(tv, 0)) { rc = -1; goto end; } \ |
| 2125 | rc = 0; \ |
| 2126 | end: \ |
| 2127 | dstr_destroy(&d); \ |
| 2128 | return (rc); \ |
| 2129 | } |
| 2130 | |
| 2131 | #define LITSTR_INT "literal signed integer" |
| 2132 | #define FOUND_INT rv->i = a->i; |
| 2133 | #define MISSING_INT if (parse_signed(&rv->i, d.buf, ei->ir, tv)) \ |
| 2134 | { rc = -1; goto end; } |
| 2135 | |
| 2136 | #define LITSTR_UINT "literal unsigned integer" |
| 2137 | #define FOUND_UINT rv->u = a->u; |
| 2138 | #define MISSING_UINT if (parse_unsigned(&rv->u, d.buf, ei->ur, tv)) \ |
| 2139 | { rc = -1; goto end; } |
| 2140 | |
| 2141 | #define LITSTR_FLT "literal floating-point number, " \ |
| 2142 | "`#-inf', `#+inf', or `#nan'" |
| 2143 | #define FOUND_FLT rv->f = a->f; |
| 2144 | #define MISSING_FLT if (parse_floating(&rv->f, 0, d.buf, ei->fi, tv)) \ |
| 2145 | { rc = -1; goto end; } |
| 2146 | |
| 2147 | #define LITSTR_PTR "`#nil'" |
| 2148 | #define FOUND_PTR rv->p = (/*unconst*/ void *)a->p; |
| 2149 | #define MISSING_PTR if (STRCMP(d.buf, ==, "#nil")) \ |
| 2150 | rv->p = 0; \ |
| 2151 | else { \ |
| 2152 | tvec_error(tv, "unknown `%s' value `%s'", \ |
| 2153 | ei->name, d.buf); \ |
| 2154 | rc = -1; goto end; \ |
| 2155 | } |
| 2156 | |
| 2157 | TVEC_MISCSLOTS(DEFPARSE_ENUM) |
| 2158 | |
| 2159 | #undef LITSTR_INT |
| 2160 | #undef FOUND_INT |
| 2161 | #undef MISSING_INT |
| 2162 | |
| 2163 | #undef LITSTR_UINT |
| 2164 | #undef FOUND_UINT |
| 2165 | #undef MISSING_UINT |
| 2166 | |
| 2167 | #undef LITSTR_FLT |
| 2168 | #undef FOUND_FLT |
| 2169 | #undef MISSING_FLT |
| 2170 | |
| 2171 | #undef LITSTR_PTR |
| 2172 | #undef FOUND_PTR |
| 2173 | #undef MISSING_PTR |
| 2174 | |
| 2175 | #undef DEFPARSE_ENUM |
| 2176 | |
| 2177 | /* --- @dump_tenum@ --- * |
| 2178 | * |
| 2179 | * Arguments: @const union tvec_regval *rv@ = register value |
| 2180 | * @const struct tvec_regdef *rd@ = register definition |
| 2181 | * @unsigned style@ = output style (@TVSF_...@) |
| 2182 | * @const struct gprintf_ops *gops@, @void *gp@ = format output |
| 2183 | * |
| 2184 | * Returns: --- |
| 2185 | * |
| 2186 | * Use: Dump a register value to the format output. |
| 2187 | * |
| 2188 | * Enumeration values are dumped as their symbolic names, if |
| 2189 | * possible, with the underlying values provided as a comment |
| 2190 | * unless compact output is requested, as for the underlying |
| 2191 | * representation. A null pointer is printed as %|#nil|%; |
| 2192 | * non-null pointers are printed as %|#<TYPE PTR>|%, with the |
| 2193 | * enumeration TYPE and the raw pointer PTR printed with the |
| 2194 | * system's %|%p|% format specifier. |
| 2195 | */ |
| 2196 | |
| 2197 | |
| 2198 | #define DEFDUMP_ENUM(tag_, ty, slot) \ |
| 2199 | static void dump_##slot##enum(const union tvec_regval *rv, \ |
| 2200 | const struct tvec_regdef *rd, \ |
| 2201 | unsigned style, \ |
| 2202 | const struct gprintf_ops *gops, void *go) \ |
| 2203 | { \ |
| 2204 | const struct tvec_##slot##enuminfo *ei = rd->arg.p; \ |
| 2205 | const struct tvec_##slot##assoc *a; \ |
| 2206 | \ |
| 2207 | for (a = ei->av; a->tag; a++) \ |
| 2208 | if (rv->slot == a->slot) { \ |
| 2209 | gprintf(gops, go, "%s", a->tag); \ |
| 2210 | if (style&TVSF_COMPACT) return; \ |
| 2211 | gprintf(gops, go, " ; = "); break; \ |
| 2212 | } \ |
| 2213 | \ |
| 2214 | PRINTRAW_##tag_ \ |
| 2215 | } |
| 2216 | |
| 2217 | #define MAYBE_PRINT_EXTRA \ |
| 2218 | if (style&TVSF_COMPACT) /* nothing to do */; \ |
| 2219 | else if (!a->tag) { gprintf(gops, go, " ; = "); goto _extra; } \ |
| 2220 | else if (1) { gprintf(gops, go, " = "); goto _extra; } \ |
| 2221 | else _extra: |
| 2222 | |
| 2223 | #define PRINTRAW_INT gprintf(gops, go, "%ld", rv->i); \ |
| 2224 | MAYBE_PRINT_EXTRA { \ |
| 2225 | format_signed_hex(gops, go, rv->i); \ |
| 2226 | maybe_format_signed_char(gops, go, rv->i); \ |
| 2227 | } |
| 2228 | |
| 2229 | #define PRINTRAW_UINT gprintf(gops, go, "%lu", rv->u); \ |
| 2230 | MAYBE_PRINT_EXTRA { \ |
| 2231 | format_unsigned_hex(gops, go, rv->u); \ |
| 2232 | maybe_format_unsigned_char(gops, go, rv->u); \ |
| 2233 | } |
| 2234 | |
| 2235 | #define PRINTRAW_FLT format_floating(gops, go, rv->f); |
| 2236 | |
| 2237 | #define PRINTRAW_PTR if (!rv->p) gprintf(gops, go, "#nil"); \ |
| 2238 | else gprintf(gops, go, "#<%s %p>", ei->name, rv->p); |
| 2239 | |
| 2240 | TVEC_MISCSLOTS(DEFDUMP_ENUM) |
| 2241 | |
| 2242 | #undef PRINTRAW_INT |
| 2243 | #undef PRINTRAW_UINT |
| 2244 | #undef PRINTRAW_FLT |
| 2245 | #undef PRINTRAW_PTR |
| 2246 | |
| 2247 | #undef MAYBE_PRINT_EXTRA |
| 2248 | #undef DEFDUMP_ENUM |
| 2249 | |
| 2250 | /* Enumeration type definitions. */ |
| 2251 | #define DEFTY_ENUM(tag, ty, slot) \ |
| 2252 | const struct tvec_regty tvty_##slot##enum = { \ |
| 2253 | init_##slot##enum, trivial_release, eq_##slot##enum, \ |
| 2254 | tobuf_##slot##enum, frombuf_##slot##enum, \ |
| 2255 | parse_##slot##enum, dump_##slot##enum \ |
| 2256 | }; |
| 2257 | TVEC_MISCSLOTS(DEFTY_ENUM) |
| 2258 | #undef DEFTY_ENUM |
| 2259 | |
| 2260 | /* Predefined enumeration types. */ |
| 2261 | static const struct tvec_iassoc bool_assoc[] = { |
| 2262 | { "nil", 0 }, |
| 2263 | { "false", 0 }, |
| 2264 | { "f", 0 }, |
| 2265 | { "no", 0 }, |
| 2266 | { "n", 0 }, |
| 2267 | { "off", 0 }, |
| 2268 | |
| 2269 | { "t", 1 }, |
| 2270 | { "true", 1 }, |
| 2271 | { "yes", 1 }, |
| 2272 | { "y", 1 }, |
| 2273 | { "on", 1 }, |
| 2274 | |
| 2275 | TVEC_ENDENUM |
| 2276 | }; |
| 2277 | |
| 2278 | const struct tvec_ienuminfo tvenum_bool = |
| 2279 | { "bool", bool_assoc, &tvrange_int }; |
| 2280 | |
| 2281 | static const struct tvec_iassoc cmp_assoc[] = { |
| 2282 | { "<", -1 }, |
| 2283 | { "less", -1 }, |
| 2284 | { "lt", -1 }, |
| 2285 | |
| 2286 | { "=", 0 }, |
| 2287 | { "equal", 0 }, |
| 2288 | { "eq", 0 }, |
| 2289 | |
| 2290 | { ">", +1 }, |
| 2291 | { "greater", +1 }, |
| 2292 | { "gt", +1 }, |
| 2293 | |
| 2294 | TVEC_ENDENUM |
| 2295 | }; |
| 2296 | |
| 2297 | const struct tvec_ienuminfo tvenum_cmp = |
| 2298 | { "cmp", cmp_assoc, &tvrange_int }; |
| 2299 | |
| 2300 | /* --- @tvec_claimeq_tenum@ --- * |
| 2301 | * |
| 2302 | * Arguments: @struct tvec_state *tv@ = test-vector state |
| 2303 | * @const struct tvec_typeenuminfo *ei@ = enumeration type info |
| 2304 | * @ty t0, t1@ = two values |
| 2305 | * @const char *file@, @unsigned @lno@ = calling file and line |
| 2306 | * @const char *expr@ = the expression to quote on failure |
| 2307 | * |
| 2308 | * Returns: Nonzero if @t0@ and @t1@ are equal, otherwise zero. |
| 2309 | * |
| 2310 | * Use: Check that values of @t0@ and @t1@ are equal. As for |
| 2311 | * @tvec_claim@ above, a test case is automatically begun and |
| 2312 | * ended if none is already underway. If the values are |
| 2313 | * unequal, then @tvec_fail@ is called, quoting @expr@, and the |
| 2314 | * mismatched values are dumped: @t0@ is printed as the output |
| 2315 | * value and @t1@ is printed as the input reference. |
| 2316 | */ |
| 2317 | |
| 2318 | #define DEFCLAIM(tag, ty, slot) \ |
| 2319 | int tvec_claimeq_##slot##enum \ |
| 2320 | (struct tvec_state *tv, \ |
| 2321 | const struct tvec_##slot##enuminfo *ei, ty e0, ty e1, \ |
| 2322 | const char *file, unsigned lno, const char *expr) \ |
| 2323 | { \ |
| 2324 | union tvec_misc arg; \ |
| 2325 | \ |
| 2326 | arg.p = ei; \ |
| 2327 | tv->out[0].v.slot = GET_##tag(e0); \ |
| 2328 | tv->in[0].v.slot = GET_##tag(e1); \ |
| 2329 | return (tvec_claimeq(tv, &tvty_##slot##enum, &arg, \ |
| 2330 | file, lno, expr)); \ |
| 2331 | } |
| 2332 | #define GET_INT(e) (e) |
| 2333 | #define GET_UINT(e) (e) |
| 2334 | #define GET_FLT(e) (e) |
| 2335 | #define GET_PTR(e) ((/*unconst*/ void *)(e)) |
| 2336 | TVEC_MISCSLOTS(DEFCLAIM) |
| 2337 | #undef DEFCLAIM |
| 2338 | #undef GET_INT |
| 2339 | #undef GET_UINT |
| 2340 | #undef GET_FLT |
| 2341 | #undef GET_PTR |
| 2342 | |
| 2343 | /*----- Flag types --------------------------------------------------------*/ |
| 2344 | |
| 2345 | /* Flag types are initialized, compared, and serialized as unsigned |
| 2346 | * integers. |
| 2347 | */ |
| 2348 | |
| 2349 | /* --- @parse_flags@ --- * |
| 2350 | * |
| 2351 | * Arguments: @union tvec_regval *rv@ = register value |
| 2352 | * @const struct tvec_regdef *rd@ = register definition |
| 2353 | * @struct tvec_state *tv@ = test-vector state |
| 2354 | * |
| 2355 | * Returns: Zero on success, %$-1$% on error. |
| 2356 | * |
| 2357 | * Use: Parse a register value from an input file. |
| 2358 | * |
| 2359 | * The input syntax is a sequence of items separated by `|' |
| 2360 | * signs. Each item may be the symbolic name of a field value, |
| 2361 | * or a literal unsigned integer. The masks associated with the |
| 2362 | * given symbolic names must be disjoint. The resulting |
| 2363 | * numerical value is simply the bitwise OR of the given values. |
| 2364 | */ |
| 2365 | |
| 2366 | static int parse_flags(union tvec_regval *rv, const struct tvec_regdef *rd, |
| 2367 | struct tvec_state *tv) |
| 2368 | { |
| 2369 | const struct tvec_flaginfo *fi = rd->arg.p; |
| 2370 | const struct tvec_flag *f; |
| 2371 | unsigned long m = 0, v = 0, t; |
| 2372 | dstr d = DSTR_INIT; |
| 2373 | int ch, rc; |
| 2374 | |
| 2375 | for (;;) { |
| 2376 | |
| 2377 | /* Read the next item. */ |
| 2378 | DRESET(&d); |
| 2379 | if (tvec_readword(tv, &d, 0, "|;", "flag name or integer")) |
| 2380 | { rc = -1; goto end; } |
| 2381 | |
| 2382 | /* Try to find a matching entry in the table. */ |
| 2383 | for (f = fi->fv; f->tag; f++) |
| 2384 | if (STRCMP(f->tag, ==, d.buf)) { |
| 2385 | if (m&f->m) |
| 2386 | { tvec_error(tv, "colliding flag setting"); rc = -1; goto end; } |
| 2387 | else |
| 2388 | { m |= f->m; v |= f->v; goto next; } |
| 2389 | } |
| 2390 | |
| 2391 | /* Otherwise, try to parse it as a raw integer. */ |
| 2392 | if (parse_unsigned(&t, d.buf, fi->range, tv)) |
| 2393 | { rc = -1; goto end; } |
| 2394 | v |= t; |
| 2395 | |
| 2396 | next: |
| 2397 | /* Advance to the next token. If it's a separator then consume it, and |
| 2398 | * go round again. Otherwise we stop here. |
| 2399 | */ |
| 2400 | if (tvec_nexttoken(tv)) break; |
| 2401 | ch = getc(tv->fp); |
| 2402 | if (ch != '|') { tvec_syntax(tv, ch, "`|'"); rc = -1; goto end; } |
| 2403 | if (tvec_nexttoken(tv)) |
| 2404 | { tvec_syntax(tv, '\n', "flag name or integer"); rc = -1; goto end; } |
| 2405 | } |
| 2406 | |
| 2407 | /* Done. */ |
| 2408 | rv->u = v; rc = 0; |
| 2409 | end: |
| 2410 | dstr_destroy(&d); |
| 2411 | return (rc); |
| 2412 | } |
| 2413 | |
| 2414 | /* --- @dump_flags@ --- * |
| 2415 | * |
| 2416 | * Arguments: @const union tvec_regval *rv@ = register value |
| 2417 | * @const struct tvec_regdef *rd@ = register definition |
| 2418 | * @unsigned style@ = output style (@TVSF_...@) |
| 2419 | * @const struct gprintf_ops *gops@, @void *gp@ = format output |
| 2420 | * |
| 2421 | * Returns: --- |
| 2422 | * |
| 2423 | * Use: Dump a register value to the format output. |
| 2424 | * |
| 2425 | * The table of symbolic names and their associated values and |
| 2426 | * masks is repeatedly scanned, in order, to find disjoint |
| 2427 | * matches -- i.e., entries whose value matches the target value |
| 2428 | * in the bit positions indicated by the mask, and whose mask |
| 2429 | * doesn't overlap with any previously found matches; the names |
| 2430 | * are then output, separated by `|'. Any remaining nonzero |
| 2431 | * bits not covered by any of the matching masks are output as a |
| 2432 | * single literal integer, in hex. |
| 2433 | * |
| 2434 | * Unless compact output is requested, or no symbolic names were |
| 2435 | * found, the raw numeric value is also printed in hex, as a |
| 2436 | * comment. |
| 2437 | */ |
| 2438 | |
| 2439 | static void dump_flags(const union tvec_regval *rv, |
| 2440 | const struct tvec_regdef *rd, |
| 2441 | unsigned style, |
| 2442 | const struct gprintf_ops *gops, void *go) |
| 2443 | { |
| 2444 | const struct tvec_flaginfo *fi = rd->arg.p; |
| 2445 | const struct tvec_flag *f; |
| 2446 | unsigned long m = ~0ul, v = rv->u; |
| 2447 | const char *sep; |
| 2448 | |
| 2449 | for (f = fi->fv, sep = ""; f->tag; f++) |
| 2450 | if ((m&f->m) && (v&f->m) == f->v) { |
| 2451 | gprintf(gops, go, "%s%s", sep, f->tag); m &= ~f->m; |
| 2452 | sep = style&TVSF_COMPACT ? "|" : " | "; |
| 2453 | } |
| 2454 | |
| 2455 | if (v&m) gprintf(gops, go, "%s0x%0*lx", sep, hex_width(v), v&m); |
| 2456 | |
| 2457 | if (m != ~0ul && !(style&TVSF_COMPACT)) |
| 2458 | gprintf(gops, go, " ; = 0x%0*lx", hex_width(rv->u), rv->u); |
| 2459 | } |
| 2460 | |
| 2461 | /* Flags type definition. */ |
| 2462 | const struct tvec_regty tvty_flags = { |
| 2463 | init_uint, trivial_release, eq_uint, |
| 2464 | tobuf_uint, frombuf_uint, |
| 2465 | parse_flags, dump_flags |
| 2466 | }; |
| 2467 | |
| 2468 | /* --- @tvec_claimeq_flags@ --- * |
| 2469 | * |
| 2470 | * Arguments: @struct tvec_state *tv@ = test-vector state |
| 2471 | * @const struct tvec_flaginfo *fi@ = flags type info |
| 2472 | * @unsigned long f0, f1@ = two values |
| 2473 | * @const char *file@, @unsigned @lno@ = calling file and line |
| 2474 | * @const char *expr@ = the expression to quote on failure |
| 2475 | * |
| 2476 | * Returns: Nonzero if @f0@ and @f1@ are equal, otherwise zero. |
| 2477 | * |
| 2478 | * Use: Check that values of @f0@ and @f1@ are equal. As for |
| 2479 | * @tvec_claim@ above, a test case is automatically begun and |
| 2480 | * ended if none is already underway. If the values are |
| 2481 | * unequal, then @tvec_fail@ is called, quoting @expr@, and the |
| 2482 | * mismatched values are dumped: @f0@ is printed as the output |
| 2483 | * value and @f1@ is printed as the input reference. |
| 2484 | */ |
| 2485 | |
| 2486 | int tvec_claimeq_flags(struct tvec_state *tv, |
| 2487 | const struct tvec_flaginfo *fi, |
| 2488 | unsigned long f0, unsigned long f1, |
| 2489 | const char *file, unsigned lno, const char *expr) |
| 2490 | { |
| 2491 | union tvec_misc arg; |
| 2492 | |
| 2493 | arg.p = fi; tv->out[0].v.u = f0; tv->in[0].v.u = f1; |
| 2494 | return (tvec_claimeq(tv, &tvty_flags, &arg, file, lno, expr)); |
| 2495 | } |
| 2496 | |
| 2497 | /*----- Characters --------------------------------------------------------*/ |
| 2498 | |
| 2499 | /* Character values are initialized and compared as signed integers. */ |
| 2500 | |
| 2501 | /* --- @tobuf_char@ --- * |
| 2502 | * |
| 2503 | * Arguments: @buf *b@ = buffer |
| 2504 | * @const union tvec_regval *rv@ = register value |
| 2505 | * @const struct tvec_regdef *rd@ = register definition |
| 2506 | * |
| 2507 | * Returns: Zero on success, %$-1$% on failure. |
| 2508 | * |
| 2509 | * Use: Serialize a register value to a buffer. |
| 2510 | * |
| 2511 | * Character values are serialized as little-endian 32-bit |
| 2512 | * unsigned integers, with %|EOF|% serialized as all-bits-set. |
| 2513 | */ |
| 2514 | |
| 2515 | static int tobuf_char(buf *b, const union tvec_regval *rv, |
| 2516 | const struct tvec_regdef *rd) |
| 2517 | { |
| 2518 | uint32 u; |
| 2519 | |
| 2520 | if (0 <= rv->i && rv->i <= UCHAR_MAX) u = rv->i; |
| 2521 | else if (rv->i == EOF) u = MASK32; |
| 2522 | else { buf_break(b); return (-1); } |
| 2523 | return (buf_putu32l(b, u)); |
| 2524 | } |
| 2525 | |
| 2526 | /* --- @frombuf_char@ --- * |
| 2527 | * |
| 2528 | * Arguments: @buf *b@ = buffer |
| 2529 | * @union tvec_regval *rv@ = register value |
| 2530 | * @const struct tvec_regdef *rd@ = register definition |
| 2531 | * |
| 2532 | * Returns: Zero on success, %$-1$% on failure. |
| 2533 | * |
| 2534 | * Use: Deserialize a register value from a buffer. |
| 2535 | * |
| 2536 | * Character values are serialized as little-endian 32-bit |
| 2537 | * unsigned integers, with %|EOF|% serialized as all-bits-set. |
| 2538 | */ |
| 2539 | |
| 2540 | static int frombuf_char(buf *b, union tvec_regval *rv, |
| 2541 | const struct tvec_regdef *rd) |
| 2542 | { |
| 2543 | uint32 u; |
| 2544 | |
| 2545 | if (buf_getu32l(b, &u)) return (-1); |
| 2546 | if (0 <= u && u <= UCHAR_MAX) rv->i = u; |
| 2547 | else if (u == MASK32) rv->i = EOF; |
| 2548 | else { buf_break(b); return (-1); } |
| 2549 | return (0); |
| 2550 | } |
| 2551 | |
| 2552 | /* --- @parse_char@ --- * |
| 2553 | * |
| 2554 | * Arguments: @union tvec_regval *rv@ = register value |
| 2555 | * @const struct tvec_regdef *rd@ = register definition |
| 2556 | * @struct tvec_state *tv@ = test-vector state |
| 2557 | * |
| 2558 | * Returns: Zero on success, %$-1$% on error. |
| 2559 | * |
| 2560 | * Use: Parse a register value from an input file. |
| 2561 | * |
| 2562 | * A character value can be given by symbolic name, with a |
| 2563 | * leading `%|#|%'; or a character or `%|\|%'-escape sequence, |
| 2564 | * optionally in single quotes. |
| 2565 | * |
| 2566 | * The following escape sequences and character names are |
| 2567 | * recognized. |
| 2568 | * |
| 2569 | * * `%|#eof|%' is the special end-of-file marker. |
| 2570 | * |
| 2571 | * * `%|#nul|%' is the NUL character, sometimes used to |
| 2572 | * terminate strings. |
| 2573 | * |
| 2574 | * * `%|bell|%', `%|bel|%', `%|ding|%', or `%|\a|%' is the BEL |
| 2575 | * character used to ring the terminal bell (or do some other |
| 2576 | * thing to attract the user's attention). |
| 2577 | * |
| 2578 | * * %|#backspace|%, %|#bs|%, or %|\b|% is the backspace |
| 2579 | * character, used to move the cursor backwords by one cell. |
| 2580 | * |
| 2581 | * * %|#escape|% %|#esc|%, or%|\e|% is the escape character, |
| 2582 | * used to introduce special terminal commands. |
| 2583 | * |
| 2584 | * * %|#formfeed|%, %|#ff|%, or %|\f|% is the formfeed |
| 2585 | * character, used to separate pages of text. |
| 2586 | * |
| 2587 | * * %|#newline|%, %|#linefeed|%, %|#lf|%, %|#nl|%, or %|\n|% is |
| 2588 | * the newline character, used to terminate lines of text or |
| 2589 | * advance the cursor to the next line (perhaps without |
| 2590 | * returning it to the start of the line). |
| 2591 | * |
| 2592 | * * %|#return|%, %|#carriage-return|%, %|#cr|%, or %|\r|% is |
| 2593 | * the carriage-return character, used to return the cursor to |
| 2594 | * the start of the line. |
| 2595 | * |
| 2596 | * * %|#tab|%, %|#horizontal-tab|%, %|#ht|%, or %|\t|% is the |
| 2597 | * tab character, used to advance the cursor to the next tab |
| 2598 | * stop on the current line. |
| 2599 | * |
| 2600 | * * %|#vertical-tab|%, %|#vt|%, %|\v|% is the vertical tab |
| 2601 | * character. |
| 2602 | * |
| 2603 | * * %|#space|%, %|#spc|% is the space character. |
| 2604 | * |
| 2605 | * * %|#delete|%, %|#del|% is the delete character, used to |
| 2606 | * erase the most recent character. |
| 2607 | * |
| 2608 | * * %|\'|% is the single-quote character. |
| 2609 | * |
| 2610 | * * %|\\|% is the backslash character. |
| 2611 | * |
| 2612 | * * %|\"|% is the double-quote character. |
| 2613 | * |
| 2614 | * * %|\NNN|% or %|\{NNN}|% is the character with code NNN in |
| 2615 | * octal. The NNN may be up to three digits long. |
| 2616 | * |
| 2617 | * * %|\xNN|% or %|\x{NN}|% is the character with code NNN in |
| 2618 | * hexadecimal. |
| 2619 | */ |
| 2620 | |
| 2621 | static int parse_char(union tvec_regval *rv, const struct tvec_regdef *rd, |
| 2622 | struct tvec_state *tv) |
| 2623 | { |
| 2624 | dstr d = DSTR_INIT; |
| 2625 | int ch, rc; |
| 2626 | unsigned f = 0; |
| 2627 | #define f_quote 1u |
| 2628 | |
| 2629 | /* Inspect the character to see what we're up against. */ |
| 2630 | ch = getc(tv->fp); |
| 2631 | |
| 2632 | if (ch == '#') { |
| 2633 | /* It looks like a special token. Push the `%|#|%' back and fetch the |
| 2634 | * whole word. If there's just the `%|#|%' after all, then treat it as |
| 2635 | * literal. |
| 2636 | */ |
| 2637 | |
| 2638 | ungetc(ch, tv->fp); |
| 2639 | if (tvec_readword(tv, &d, 0, ";", "character name")) |
| 2640 | { rc = -1; goto end; } |
| 2641 | if (STRCMP(d.buf, !=, "#")) { |
| 2642 | if (read_charname(&ch, d.buf, RCF_EOFOK)) { |
| 2643 | rc = tvec_error(tv, "unknown character name `%s'", d.buf); |
| 2644 | goto end; |
| 2645 | } |
| 2646 | if (tvec_flushtoeol(tv, 0)) { rc = -1; goto end; } |
| 2647 | rv->i = ch; rc = 0; goto end; |
| 2648 | } |
| 2649 | } |
| 2650 | |
| 2651 | /* If this is a single quote then we expect to see a matching one later, |
| 2652 | * and we should process backslash escapes. Get the next character and see |
| 2653 | * what happens. |
| 2654 | */ |
| 2655 | if (ch == '\'') { f |= f_quote; ch = getc(tv->fp); } |
| 2656 | |
| 2657 | /* Main character dispatch. */ |
| 2658 | switch (ch) { |
| 2659 | |
| 2660 | case ';': |
| 2661 | /* Unquoted, semicolon begins a comment. */ |
| 2662 | if (!(f&f_quote)) { rc = tvec_syntax(tv, ch, "character"); goto end; } |
| 2663 | else goto plain; |
| 2664 | |
| 2665 | case '\n': |
| 2666 | /* A newline. If we saw a single quote, then treat that as literal. |
| 2667 | * Otherwise this is an error. |
| 2668 | */ |
| 2669 | if (!(f&f_quote)) goto nochar; |
| 2670 | else { f &= ~f_quote; ungetc(ch, tv->fp); ch = '\''; goto plain; } |
| 2671 | |
| 2672 | case EOF: |
| 2673 | /* End-of-file. Similar to newline, but with slightly different |
| 2674 | * effects on the parse state. |
| 2675 | */ |
| 2676 | if (!(f&f_quote)) goto nochar; |
| 2677 | else { f &= ~f_quote; ch = '\''; goto plain; } |
| 2678 | |
| 2679 | case '\'': nochar: |
| 2680 | /* A single quote. This must be the second of a pair, and there should |
| 2681 | * have been a character or escape sequence between them. |
| 2682 | */ |
| 2683 | rc = tvec_syntax(tv, ch, "character"); goto end; |
| 2684 | |
| 2685 | case '\\': |
| 2686 | /* A backslash. Read a character escape. */ |
| 2687 | if (read_charesc(&ch, tv)) return (-1); |
| 2688 | |
| 2689 | default: plain: |
| 2690 | /* Anything else. Treat as literal. */ |
| 2691 | rv->i = ch; break; |
| 2692 | } |
| 2693 | |
| 2694 | /* If we saw an opening quote, then expect the closing quote. */ |
| 2695 | if (f&f_quote) { |
| 2696 | ch = getc(tv->fp); |
| 2697 | if (ch != '\'') { rc = tvec_syntax(tv, ch, "`''"); goto end; } |
| 2698 | } |
| 2699 | |
| 2700 | /* Done. */ |
| 2701 | if (tvec_flushtoeol(tv, 0)) { rc = -1; goto end; } |
| 2702 | rc = 0; |
| 2703 | end: |
| 2704 | dstr_destroy(&d); |
| 2705 | return (rc); |
| 2706 | |
| 2707 | #undef f_quote |
| 2708 | } |
| 2709 | |
| 2710 | /* --- @dump_char@ --- * |
| 2711 | * |
| 2712 | * Arguments: @const union tvec_regval *rv@ = register value |
| 2713 | * @const struct tvec_regdef *rd@ = register definition |
| 2714 | * @unsigned style@ = output style (@TVSF_...@) |
| 2715 | * @const struct gprintf_ops *gops@, @void *gp@ = format output |
| 2716 | * |
| 2717 | * Returns: --- |
| 2718 | * |
| 2719 | * Use: Dump a register value to the format output. |
| 2720 | * |
| 2721 | * Character values are dumped as their symbolic names, if any, |
| 2722 | * or as a character or escape sequence within single quotes |
| 2723 | * (which may be omitted in compact style). If compact output |
| 2724 | * is not requested, then the single-quoted representation (for |
| 2725 | * characters dumped as symbolic names) and integer code in |
| 2726 | * decimal and hex are printed as a comment. |
| 2727 | */ |
| 2728 | |
| 2729 | static void dump_char(const union tvec_regval *rv, |
| 2730 | const struct tvec_regdef *rd, |
| 2731 | unsigned style, |
| 2732 | const struct gprintf_ops *gops, void *go) |
| 2733 | { |
| 2734 | const char *p; |
| 2735 | unsigned f = 0; |
| 2736 | #define f_semi 1u |
| 2737 | |
| 2738 | /* Print a character name if we can find one. */ |
| 2739 | p = find_charname(rv->i, (style&TVSF_COMPACT) ? CTF_SHORT : CTF_PREFER); |
| 2740 | if (p) { |
| 2741 | gprintf(gops, go, "%s", p); |
| 2742 | if (style&TVSF_COMPACT) return; |
| 2743 | else { gprintf(gops, go, " ;"); f |= f_semi; } |
| 2744 | } |
| 2745 | |
| 2746 | /* If the character isn't @EOF@ then print it as a single-quoted thing. |
| 2747 | * In compact style, see if we can omit the quotes. |
| 2748 | */ |
| 2749 | if (rv->i >= 0) { |
| 2750 | if (f&f_semi) gprintf(gops, go, " = "); |
| 2751 | switch (rv->i) { |
| 2752 | case ' ': case '\\': case '\'': quote: |
| 2753 | format_char(gops, go, rv->i); |
| 2754 | break; |
| 2755 | default: |
| 2756 | if (!(style&TVSF_COMPACT) || !isprint(rv->i)) goto quote; |
| 2757 | gprintf(gops, go, "%c", (int)rv->i); |
| 2758 | return; |
| 2759 | } |
| 2760 | } |
| 2761 | |
| 2762 | /* And the character code as an integer. */ |
| 2763 | if (!(style&TVSF_COMPACT)) { |
| 2764 | if (!(f&f_semi)) gprintf(gops, go, " ;"); |
| 2765 | gprintf(gops, go, " = %ld = ", rv->i); |
| 2766 | format_signed_hex(gops, go, rv->i); |
| 2767 | } |
| 2768 | |
| 2769 | #undef f_semi |
| 2770 | } |
| 2771 | |
| 2772 | /* Character type definition. */ |
| 2773 | const struct tvec_regty tvty_char = { |
| 2774 | init_int, trivial_release, eq_int, |
| 2775 | tobuf_char, frombuf_char, |
| 2776 | parse_char, dump_char |
| 2777 | }; |
| 2778 | |
| 2779 | /* --- @tvec_claimeq_char@ --- * |
| 2780 | * |
| 2781 | * Arguments: @struct tvec_state *tv@ = test-vector state |
| 2782 | * @int ch0, ch1@ = two character codes |
| 2783 | * @const char *file@, @unsigned @lno@ = calling file and line |
| 2784 | * @const char *expr@ = the expression to quote on failure |
| 2785 | * |
| 2786 | * Returns: Nonzero if @ch0@ and @ch1@ are equal, otherwise zero. |
| 2787 | * |
| 2788 | * Use: Check that values of @ch0@ and @ch1@ are equal. As for |
| 2789 | * @tvec_claim@ above, a test case is automatically begun and |
| 2790 | * ended if none is already underway. If the values are |
| 2791 | * unequal, then @tvec_fail@ is called, quoting @expr@, and the |
| 2792 | * mismatched values are dumped: @ch0@ is printed as the output |
| 2793 | * value and @ch1@ is printed as the input reference. |
| 2794 | */ |
| 2795 | |
| 2796 | int tvec_claimeq_char(struct tvec_state *tv, int c0, int c1, |
| 2797 | const char *file, unsigned lno, const char *expr) |
| 2798 | { |
| 2799 | tv->out[0].v.i = c0; tv->in[0].v.i = c1; |
| 2800 | return (tvec_claimeq(tv, &tvty_char, 0, file, lno, expr)); |
| 2801 | } |
| 2802 | |
| 2803 | /*----- Text and byte strings ---------------------------------------------*/ |
| 2804 | |
| 2805 | /* --- @init_text@, @init_bytes@ --- * |
| 2806 | * |
| 2807 | * Arguments: @union tvec_regval *rv@ = register value |
| 2808 | * @const struct tvec_regdef *rd@ = register definition |
| 2809 | * |
| 2810 | * Returns: --- |
| 2811 | * |
| 2812 | * Use: Initialize a register value. |
| 2813 | * |
| 2814 | * Text and binary string values are initialized with a null |
| 2815 | * pointer and zero length. |
| 2816 | */ |
| 2817 | |
| 2818 | static void init_text(union tvec_regval *rv, const struct tvec_regdef *rd) |
| 2819 | { rv->text.p = 0; rv->text.sz = 0; } |
| 2820 | |
| 2821 | static void init_bytes(union tvec_regval *rv, const struct tvec_regdef *rd) |
| 2822 | { rv->bytes.p = 0; rv->bytes.sz = 0; } |
| 2823 | |
| 2824 | /* --- @release_string@, @release_bytes@ --- * |
| 2825 | * |
| 2826 | * Arguments: @const union tvec_regval *rv@ = register value |
| 2827 | * @const struct tvec_regdef *rd@ = register definition |
| 2828 | * |
| 2829 | * Returns: --- |
| 2830 | * |
| 2831 | * Use: Release resources held by a register value. |
| 2832 | * |
| 2833 | * Text and binary string buffers are freed. |
| 2834 | */ |
| 2835 | |
| 2836 | static void release_text(union tvec_regval *rv, |
| 2837 | const struct tvec_regdef *rd) |
| 2838 | { xfree(rv->text.p); } |
| 2839 | |
| 2840 | static void release_bytes(union tvec_regval *rv, |
| 2841 | const struct tvec_regdef *rd) |
| 2842 | { xfree(rv->bytes.p); } |
| 2843 | |
| 2844 | /* --- @eq_text@, @eq_bytes@ --- * |
| 2845 | * |
| 2846 | * Arguments: @const union tvec_regval *rv0, *rv1@ = register values |
| 2847 | * @const struct tvec_regdef *rd@ = register definition |
| 2848 | * |
| 2849 | * Returns: Nonzero if the values are equal, zero if unequal |
| 2850 | * |
| 2851 | * Use: Compare register values for equality. |
| 2852 | */ |
| 2853 | |
| 2854 | static int eq_text(const union tvec_regval *rv0, |
| 2855 | const union tvec_regval *rv1, |
| 2856 | const struct tvec_regdef *rd) |
| 2857 | { |
| 2858 | return (rv0->text.sz == rv1->text.sz && |
| 2859 | (!rv0->text.sz || |
| 2860 | MEMCMP(rv0->text.p, ==, rv1->text.p, rv1->text.sz))); |
| 2861 | } |
| 2862 | |
| 2863 | static int eq_bytes(const union tvec_regval *rv0, |
| 2864 | const union tvec_regval *rv1, |
| 2865 | const struct tvec_regdef *rd) |
| 2866 | { |
| 2867 | return (rv0->bytes.sz == rv1->bytes.sz && |
| 2868 | (!rv0->bytes.sz || |
| 2869 | MEMCMP(rv0->bytes.p, ==, rv1->bytes.p, rv1->bytes.sz))); |
| 2870 | } |
| 2871 | |
| 2872 | /* --- @tobuf_text@, @tobuf_bytes@ --- * |
| 2873 | * |
| 2874 | * Arguments: @buf *b@ = buffer |
| 2875 | * @const union tvec_regval *rv@ = register value |
| 2876 | * @const struct tvec_regdef *rd@ = register definition |
| 2877 | * |
| 2878 | * Returns: Zero on success, %$-1$% on failure. |
| 2879 | * |
| 2880 | * Use: Serialize a register value to a buffer. |
| 2881 | * |
| 2882 | * Text and binary string values are serialized as a little- |
| 2883 | * endian 64-bit length %$n$% in bytes followed by %$n$% bytes |
| 2884 | * of string data. |
| 2885 | */ |
| 2886 | |
| 2887 | static int tobuf_text(buf *b, const union tvec_regval *rv, |
| 2888 | const struct tvec_regdef *rd) |
| 2889 | { return (buf_putmem64l(b, rv->text.p, rv->text.sz)); } |
| 2890 | |
| 2891 | static int tobuf_bytes(buf *b, const union tvec_regval *rv, |
| 2892 | const struct tvec_regdef *rd) |
| 2893 | { return (buf_putmem64l(b, rv->bytes.p, rv->bytes.sz)); } |
| 2894 | |
| 2895 | /* --- @frombuf_text@, @frombuf_bytes@ --- * |
| 2896 | * |
| 2897 | * Arguments: @buf *b@ = buffer |
| 2898 | * @union tvec_regval *rv@ = register value |
| 2899 | * @const struct tvec_regdef *rd@ = register definition |
| 2900 | * |
| 2901 | * Returns: Zero on success, %$-1$% on failure. |
| 2902 | * |
| 2903 | * Use: Deserialize a register value from a buffer. |
| 2904 | * |
| 2905 | * Text and binary string values are serialized as a little- |
| 2906 | * endian 64-bit length %$n$% in bytes followed by %$n$% bytes |
| 2907 | * of string data. |
| 2908 | */ |
| 2909 | |
| 2910 | static int frombuf_text(buf *b, union tvec_regval *rv, |
| 2911 | const struct tvec_regdef *rd) |
| 2912 | { |
| 2913 | const void *p; |
| 2914 | size_t sz; |
| 2915 | |
| 2916 | p = buf_getmem64l(b, &sz); if (!p) return (-1); |
| 2917 | tvec_alloctext(rv, sz); memcpy(rv->text.p, p, sz); rv->text.p[sz] = 0; |
| 2918 | return (0); |
| 2919 | } |
| 2920 | |
| 2921 | static int frombuf_bytes(buf *b, union tvec_regval *rv, |
| 2922 | const struct tvec_regdef *rd) |
| 2923 | { |
| 2924 | const void *p; |
| 2925 | size_t sz; |
| 2926 | |
| 2927 | p = buf_getmem64l(b, &sz); if (!p) return (-1); |
| 2928 | tvec_allocbytes(rv, sz); memcpy(rv->bytes.p, p, sz); |
| 2929 | return (0); |
| 2930 | } |
| 2931 | |
| 2932 | /* --- @check_string_length@ --- * |
| 2933 | * |
| 2934 | * Arguments: @size_t sz@ = found string length |
| 2935 | * @const struct tvec_urange *ur@ = acceptable range |
| 2936 | * @struct tvec_state *tv@ = test-vector state |
| 2937 | * |
| 2938 | * Returns: Zero on success, %$-1$% on error. |
| 2939 | * |
| 2940 | * Use: Checks that @sz@ is within the bounds described by @ur@, |
| 2941 | * reporting an error if not. |
| 2942 | */ |
| 2943 | |
| 2944 | static int check_string_length(size_t sz, const struct tvec_urange *ur, |
| 2945 | struct tvec_state *tv) |
| 2946 | { |
| 2947 | if (ur && (ur->min > sz || sz > ur->max)) |
| 2948 | return (tvec_error(tv, |
| 2949 | "invalid string length %lu; must be in [%lu .. %lu]", |
| 2950 | (unsigned long)sz, ur->min, ur->max)); |
| 2951 | return (0); |
| 2952 | } |
| 2953 | |
| 2954 | /* --- @parse_text@, @parse_bytes@ --- * |
| 2955 | * |
| 2956 | * Arguments: @union tvec_regval *rv@ = register value |
| 2957 | * @const struct tvec_regdef *rd@ = register definition |
| 2958 | * @struct tvec_state *tv@ = test-vector state |
| 2959 | * |
| 2960 | * Returns: Zero on success, %$-1$% on error. |
| 2961 | * |
| 2962 | * Use: Parse a register value from an input file. |
| 2963 | * |
| 2964 | * The input format for both kinds of strings is basically the |
| 2965 | * same: a `compound string', consisting of |
| 2966 | * |
| 2967 | * * single-quoted strings, which are interpreted entirely |
| 2968 | * literally, but can't contain single quotes or newlines; |
| 2969 | * |
| 2970 | * * double-quoted strings, in which `%|\|%'-escapes are |
| 2971 | * interpreted as for characters; |
| 2972 | * |
| 2973 | * * character names, marked by an initial `%|#|%' sign; |
| 2974 | * |
| 2975 | * * special tokens marked by an initial `%|!|%' sign; or |
| 2976 | * |
| 2977 | * * barewords interpreted according to the current coding |
| 2978 | * scheme. |
| 2979 | * |
| 2980 | * The special tokens are |
| 2981 | * |
| 2982 | * * `%|!bare|%', which causes subsequent sequences of |
| 2983 | * barewords to be treated as plain text; |
| 2984 | * |
| 2985 | * * `%|!hex|%', `%|!base32|%', `%|!base64|%', which cause |
| 2986 | * subsequent barewords to be decoded in the requested |
| 2987 | * manner. |
| 2988 | * |
| 2989 | * * `%|!repeat|% %$n$% %|{|% %%\textit{string}%% %|}|%', |
| 2990 | * which includes %$n$% copies of the (compound) string. |
| 2991 | * |
| 2992 | * The only difference between text and binary strings is that |
| 2993 | * the initial coding scheme is %|bare|% for text strings and |
| 2994 | * %|hex|% for binary strings. |
| 2995 | */ |
| 2996 | |
| 2997 | static int parse_text(union tvec_regval *rv, const struct tvec_regdef *rd, |
| 2998 | struct tvec_state *tv) |
| 2999 | { |
| 3000 | void *p = rv->text.p; |
| 3001 | |
| 3002 | if (read_compound_string(&p, &rv->text.sz, TVCODE_BARE, 0, tv)) |
| 3003 | return (-1); |
| 3004 | rv->text.p = p; |
| 3005 | if (check_string_length(rv->text.sz, rd->arg.p, tv)) return (-1); |
| 3006 | return (0); |
| 3007 | } |
| 3008 | |
| 3009 | static int parse_bytes(union tvec_regval *rv, const struct tvec_regdef *rd, |
| 3010 | struct tvec_state *tv) |
| 3011 | { |
| 3012 | void *p = rv->bytes.p; |
| 3013 | |
| 3014 | if (read_compound_string(&p, &rv->bytes.sz, TVCODE_HEX, 0, tv)) |
| 3015 | return (-1); |
| 3016 | rv->bytes.p = p; |
| 3017 | if (check_string_length(rv->bytes.sz, rd->arg.p, tv)) return (-1); |
| 3018 | return (0); |
| 3019 | } |
| 3020 | |
| 3021 | /* --- @dump_text@, @dump_bytes@ --- * |
| 3022 | * |
| 3023 | * Arguments: @const union tvec_regval *rv@ = register value |
| 3024 | * @const struct tvec_regdef *rd@ = register definition |
| 3025 | * @unsigned style@ = output style (@TVSF_...@) |
| 3026 | * @const struct gprintf_ops *gops@, @void *gp@ = format output |
| 3027 | * |
| 3028 | * Returns: --- |
| 3029 | * |
| 3030 | * Use: Dump a register value to the format output. |
| 3031 | * |
| 3032 | * Text string values are dumped as plain text, in double quotes |
| 3033 | * if necessary, and using backslash escape sequences for |
| 3034 | * nonprintable characters. Unless compact output is requested, |
| 3035 | * strings consisting of multiple lines are dumped with each |
| 3036 | * line of the string on a separate output line. |
| 3037 | * |
| 3038 | * Binary string values are dumped in hexadecimal. In compact |
| 3039 | * style, the output simply consists of a single block of hex |
| 3040 | * digits. Otherwise, the dump is a display consisting of |
| 3041 | * groups of hex digits, with comments showing the offset (if |
| 3042 | * the string is long enough) and the corresponding plain text. |
| 3043 | * |
| 3044 | * Empty strings are dumped as %|""|%. |
| 3045 | */ |
| 3046 | |
| 3047 | static void dump_text(const union tvec_regval *rv, |
| 3048 | const struct tvec_regdef *rd, |
| 3049 | unsigned style, |
| 3050 | const struct gprintf_ops *gops, void *go) |
| 3051 | { |
| 3052 | const unsigned char *p, *q, *l; |
| 3053 | unsigned f = 0; |
| 3054 | #define f_nonword 1u |
| 3055 | #define f_newline 2u |
| 3056 | |
| 3057 | if (!rv->text.sz) { gprintf(gops, go, "\"\""); return; } |
| 3058 | |
| 3059 | p = (const unsigned char *)rv->text.p; l = p + rv->text.sz; |
| 3060 | switch (*p) { |
| 3061 | case '!': case '#': case ';': case '"': case '\'': |
| 3062 | case '(': case '{': case '[': case ']': case '}': case ')': |
| 3063 | f |= f_nonword; break; |
| 3064 | } |
| 3065 | for (q = p; q < l; q++) |
| 3066 | if (*q == '\n' && q != l - 1) f |= f_newline; |
| 3067 | else if (!*q || !isgraph(*q) || *q == '\\') f |= f_nonword; |
| 3068 | if (f&f_newline) { gprintf(gops, go, "\n\t"); goto quote; } |
| 3069 | else if (f&f_nonword) goto quote; |
| 3070 | |
| 3071 | gops->putm(go, (const char *)p, rv->text.sz); |
| 3072 | return; |
| 3073 | |
| 3074 | quote: |
| 3075 | gprintf(gops, go, "\""); |
| 3076 | for (q = p; q < l; q++) |
| 3077 | if (!isprint(*q) || *q == '"') { |
| 3078 | if (p < q) gops->putm(go, (const char *)p, q - p); |
| 3079 | if (*q != '\n' || (style&TVSF_COMPACT)) |
| 3080 | format_charesc(gops, go, *q, FCF_BRACE); |
| 3081 | else { |
| 3082 | if (q + 1 == l) { gprintf(gops, go, "\\n\""); return; } |
| 3083 | else gprintf(gops, go, "\\n\"\n\t\""); |
| 3084 | } |
| 3085 | p = q + 1; |
| 3086 | } |
| 3087 | if (p < q) gops->putm(go, (const char *)p, q - p); |
| 3088 | gprintf(gops, go, "\""); |
| 3089 | |
| 3090 | #undef f_nonword |
| 3091 | #undef f_newline |
| 3092 | } |
| 3093 | |
| 3094 | static void dump_bytes(const union tvec_regval *rv, |
| 3095 | const struct tvec_regdef *rd, |
| 3096 | unsigned style, |
| 3097 | const struct gprintf_ops *gops, void *go) |
| 3098 | { |
| 3099 | const unsigned char *p = rv->bytes.p, *l = p + rv->bytes.sz; |
| 3100 | size_t off, sz = rv->bytes.sz; |
| 3101 | unsigned i, n; |
| 3102 | int wd; |
| 3103 | |
| 3104 | if (!sz) { |
| 3105 | gprintf(gops, go, style&TVSF_COMPACT ? "\"\"" : "\"\" ; empty"); |
| 3106 | return; |
| 3107 | } |
| 3108 | |
| 3109 | if (style&TVSF_COMPACT) { |
| 3110 | while (p < l) gprintf(gops, go, "%02x", *p++); |
| 3111 | return; |
| 3112 | } |
| 3113 | |
| 3114 | if (sz > 16) gprintf(gops, go, "\n\t"); |
| 3115 | |
| 3116 | off = 0; wd = hex_width(sz); |
| 3117 | while (p < l) { |
| 3118 | if (l - p < 16) n = l - p; |
| 3119 | else n = 16; |
| 3120 | |
| 3121 | for (i = 0; i < n; i++) { |
| 3122 | if (i < n) gprintf(gops, go, "%02x", p[i]); |
| 3123 | else gprintf(gops, go, " "); |
| 3124 | if (i < n - 1 && i%4 == 3) gprintf(gops, go, " "); |
| 3125 | } |
| 3126 | gprintf(gops, go, " ; "); |
| 3127 | if (sz > 16) gprintf(gops, go, "[%0*lx] ", wd, (unsigned long)off); |
| 3128 | for (i = 0; i < n; i++) |
| 3129 | gprintf(gops, go, "%c", isprint(p[i]) ? p[i] : '.'); |
| 3130 | p += n; off += n; |
| 3131 | if (p < l) gprintf(gops, go, "\n\t"); |
| 3132 | } |
| 3133 | } |
| 3134 | |
| 3135 | /* Text and byte string type definitions. */ |
| 3136 | const struct tvec_regty tvty_text = { |
| 3137 | init_text, release_text, eq_text, |
| 3138 | tobuf_text, frombuf_text, |
| 3139 | parse_text, dump_text |
| 3140 | }; |
| 3141 | const struct tvec_regty tvty_bytes = { |
| 3142 | init_bytes, release_bytes, eq_bytes, |
| 3143 | tobuf_bytes, frombuf_bytes, |
| 3144 | parse_bytes, dump_bytes |
| 3145 | }; |
| 3146 | |
| 3147 | /* --- @tvec_claimeq_text@ --- * |
| 3148 | * |
| 3149 | * Arguments: @struct tvec_state *tv@ = test-vector state |
| 3150 | * @const char *p0@, @size_t sz0@ = first string with length |
| 3151 | * @const char *p1@, @size_t sz1@ = second string with length |
| 3152 | * @const char *file@, @unsigned @lno@ = calling file and line |
| 3153 | * @const char *expr@ = the expression to quote on failure |
| 3154 | * |
| 3155 | * Returns: Nonzero if the strings at @p0@ and @p1@ are equal, otherwise |
| 3156 | * zero. |
| 3157 | * |
| 3158 | * Use: Check that strings at @p0@ and @p1@ are equal. As for |
| 3159 | * @tvec_claim@ above, a test case is automatically begun and |
| 3160 | * ended if none is already underway. If the values are |
| 3161 | * unequal, then @tvec_fail@ is called, quoting @expr@, and the |
| 3162 | * mismatched values are dumped: @p0@ is printed as the output |
| 3163 | * value and @p1@ is printed as the input reference. |
| 3164 | */ |
| 3165 | |
| 3166 | int tvec_claimeq_text(struct tvec_state *tv, |
| 3167 | const char *p0, size_t sz0, |
| 3168 | const char *p1, size_t sz1, |
| 3169 | const char *file, unsigned lno, const char *expr) |
| 3170 | { |
| 3171 | tv->out[0].v.text.p = (/*unconst*/ char *)p0; tv->out[0].v.text.sz = sz0; |
| 3172 | tv->in[0].v.text.p =(/*unconst*/ char *) p1; tv->in[0].v.text.sz = sz1; |
| 3173 | return (tvec_claimeq(tv, &tvty_text, 0, file, lno, expr)); |
| 3174 | } |
| 3175 | |
| 3176 | /* --- @tvec_claimeq_textz@ --- * |
| 3177 | * |
| 3178 | * Arguments: @struct tvec_state *tv@ = test-vector state |
| 3179 | * @const char *p0, *p1@ = two strings to compare |
| 3180 | * @const char *file@, @unsigned @lno@ = calling file and line |
| 3181 | * @const char *expr@ = the expression to quote on failure |
| 3182 | * |
| 3183 | * Returns: Nonzero if the strings at @p0@ and @p1@ are equal, otherwise |
| 3184 | * zero. |
| 3185 | * |
| 3186 | * Use: Check that strings at @p0@ and @p1@ are equal, as for |
| 3187 | * @tvec_claimeq_string@, except that the strings are assumed |
| 3188 | * null-terminated, so their lengths don't need to be supplied |
| 3189 | * explicitly. |
| 3190 | */ |
| 3191 | |
| 3192 | int tvec_claimeq_textz(struct tvec_state *tv, |
| 3193 | const char *p0, const char *p1, |
| 3194 | const char *file, unsigned lno, const char *expr) |
| 3195 | { |
| 3196 | tv->out[0].v.text.p = (/*unconst*/ char *)p0; |
| 3197 | tv->out[0].v.text.sz = strlen(p0); |
| 3198 | tv->in[0].v.text.p = (/*unconst*/ char *)p1; |
| 3199 | tv->in[0].v.text.sz = strlen(p1); |
| 3200 | return (tvec_claimeq(tv, &tvty_text, 0, file, lno, expr)); |
| 3201 | } |
| 3202 | |
| 3203 | /* --- @tvec_claimeq_bytes@ --- * |
| 3204 | * |
| 3205 | * Arguments: @struct tvec_state *tv@ = test-vector state |
| 3206 | * @const void *p0@, @size_t sz0@ = first string with length |
| 3207 | * @const void *p1@, @size_t sz1@ = second string with length |
| 3208 | * @const char *file@, @unsigned @lno@ = calling file and line |
| 3209 | * @const char *expr@ = the expression to quote on failure |
| 3210 | * |
| 3211 | * Returns: Nonzero if the strings at @p0@ and @p1@ are equal, otherwise |
| 3212 | * zero. |
| 3213 | * |
| 3214 | * Use: Check that binary strings at @p0@ and @p1@ are equal. As for |
| 3215 | * @tvec_claim@ above, a test case is automatically begun and |
| 3216 | * ended if none is already underway. If the values are |
| 3217 | * unequal, then @tvec_fail@ is called, quoting @expr@, and the |
| 3218 | * mismatched values are dumped: @p0@ is printed as the output |
| 3219 | * value and @p1@ is printed as the input reference. |
| 3220 | */ |
| 3221 | |
| 3222 | int tvec_claimeq_bytes(struct tvec_state *tv, |
| 3223 | const void *p0, size_t sz0, |
| 3224 | const void *p1, size_t sz1, |
| 3225 | const char *file, unsigned lno, const char *expr) |
| 3226 | { |
| 3227 | tv->out[0].v.bytes.p = (/*unconst*/ void *)p0; |
| 3228 | tv->out[0].v.bytes.sz = sz0; |
| 3229 | tv->in[0].v.bytes.p = (/*unconst*/ void *)p1; |
| 3230 | tv->in[0].v.bytes.sz = sz1; |
| 3231 | return (tvec_claimeq(tv, &tvty_bytes, 0, file, lno, expr)); |
| 3232 | } |
| 3233 | |
| 3234 | /* --- @tvec_alloctext@, @tvec_allocbytes@ --- * |
| 3235 | * |
| 3236 | * Arguments: @union tvec_regval *rv@ = register value |
| 3237 | * @size_t sz@ = required size |
| 3238 | * |
| 3239 | * Returns: --- |
| 3240 | * |
| 3241 | * Use: Allocated space in a text or binary string register. If the |
| 3242 | * current register size is sufficient, its buffer is left |
| 3243 | * alone; otherwise, the old buffer, if any, is freed and a |
| 3244 | * fresh buffer allocated. These functions are not intended to |
| 3245 | * be used to adjust a buffer repeatedly, e.g., while building |
| 3246 | * output incrementally: (a) they will perform badly, and (b) |
| 3247 | * the old buffer contents are simply discarded if reallocation |
| 3248 | * is necessary. Instead, use a @dbuf@ or @dstr@. |
| 3249 | * |
| 3250 | * The @tvec_alloctext@ function sneakily allocates an extra |
| 3251 | * byte for a terminating zero. The @tvec_allocbytes@ function |
| 3252 | * doesn't do this. |
| 3253 | */ |
| 3254 | |
| 3255 | void tvec_alloctext(union tvec_regval *rv, size_t sz) |
| 3256 | { |
| 3257 | if (rv->text.sz <= sz) { xfree(rv->text.p); rv->text.p = xmalloc(sz + 1); } |
| 3258 | rv->text.sz = sz; |
| 3259 | } |
| 3260 | |
| 3261 | void tvec_allocbytes(union tvec_regval *rv, size_t sz) |
| 3262 | { |
| 3263 | if (rv->bytes.sz < sz) { xfree(rv->bytes.p); rv->bytes.p = xmalloc(sz); } |
| 3264 | rv->bytes.sz = sz; |
| 3265 | } |
| 3266 | |
| 3267 | /*----- Buffer type -------------------------------------------------------*/ |
| 3268 | |
| 3269 | /* --- @init_buffer@ --- * |
| 3270 | * |
| 3271 | * Arguments: @union tvec_regval *rv@ = register value |
| 3272 | * @const struct tvec_regdef *rd@ = register definition |
| 3273 | * |
| 3274 | * Returns: --- |
| 3275 | * |
| 3276 | * Use: Initialize a register value. |
| 3277 | * |
| 3278 | * Buffer values values are initialized with a null pointer, |
| 3279 | * zero length, and zero residue, modulus, and offset. |
| 3280 | */ |
| 3281 | |
| 3282 | static void init_buffer(union tvec_regval *rv, const struct tvec_regdef *rd) |
| 3283 | { rv->buf.p = 0; rv->buf.sz = rv->buf.a = rv->buf.m = rv->buf.off = 0; } |
| 3284 | |
| 3285 | /* --- @release_buffer@, @release_bytes@ --- * |
| 3286 | * |
| 3287 | * Arguments: @const union tvec_regval *rv@ = register value |
| 3288 | * @const struct tvec_regdef *rd@ = register definition |
| 3289 | * |
| 3290 | * Returns: --- |
| 3291 | * |
| 3292 | * Use: Release resources held by a register value. |
| 3293 | * |
| 3294 | * Buffers are freed. |
| 3295 | */ |
| 3296 | |
| 3297 | static void release_buffer(union tvec_regval *rv, |
| 3298 | const struct tvec_regdef *rd) |
| 3299 | { if (rv->buf.p) xfree(rv->buf.p - rv->buf.off); } |
| 3300 | |
| 3301 | /* --- @eq_buffer@ --- * |
| 3302 | * |
| 3303 | * Arguments: @const union tvec_regval *rv0, *rv1@ = register values |
| 3304 | * @const struct tvec_regdef *rd@ = register definition |
| 3305 | * |
| 3306 | * Returns: Nonzero if the values are equal, zero if unequal |
| 3307 | * |
| 3308 | * Use: Compare register values for equality. |
| 3309 | * |
| 3310 | * Buffer values are equal if and only if their sizes and |
| 3311 | * alignment parameters are equal; their contents are |
| 3312 | * %%\emph{not}%% compared. |
| 3313 | */ |
| 3314 | |
| 3315 | static int eq_buffer(const union tvec_regval *rv0, |
| 3316 | const union tvec_regval *rv1, |
| 3317 | const struct tvec_regdef *rd) |
| 3318 | { |
| 3319 | return (rv0->buf.sz == rv1->buf.sz && |
| 3320 | rv0->buf.a == rv1->buf.a && |
| 3321 | rv0->buf.m == rv1->buf.m); |
| 3322 | } |
| 3323 | |
| 3324 | /* --- @tobuf_buffer@ --- * |
| 3325 | * |
| 3326 | * Arguments: @buf *b@ = buffer |
| 3327 | * @const union tvec_regval *rv@ = register value |
| 3328 | * @const struct tvec_regdef *rd@ = register definition |
| 3329 | * |
| 3330 | * Returns: Zero on success, %$-1$% on failure. |
| 3331 | * |
| 3332 | * Use: Serialize a register value to a buffer. |
| 3333 | * |
| 3334 | * Buffer values are serialized as their lengths, residues, and |
| 3335 | * moduli, as unsigned integers. |
| 3336 | */ |
| 3337 | |
| 3338 | static int tobuf_buffer(buf *b, const union tvec_regval *rv, |
| 3339 | const struct tvec_regdef *rd) |
| 3340 | { |
| 3341 | return (unsigned_to_buf(b, rv->buf.sz) || |
| 3342 | unsigned_to_buf(b, rv->buf.a) || |
| 3343 | unsigned_to_buf(b, rv->buf.m)); |
| 3344 | } |
| 3345 | |
| 3346 | /* --- @frombuf_buffer@ --- * |
| 3347 | * |
| 3348 | * Arguments: @buf *b@ = buffer |
| 3349 | * @union tvec_regval *rv@ = register value |
| 3350 | * @const struct tvec_regdef *rd@ = register definition |
| 3351 | * |
| 3352 | * Returns: Zero on success, %$-1$% on failure. |
| 3353 | * |
| 3354 | * Use: Deserialize a register value from a buffer. |
| 3355 | * |
| 3356 | * Buffer values are serialized as just their lengths, as |
| 3357 | * unsigned integers. The buffer is allocated on |
| 3358 | * deserialization and filled with a distinctive pattern. |
| 3359 | */ |
| 3360 | |
| 3361 | static int frombuf_buffer(buf *b, union tvec_regval *rv, |
| 3362 | const struct tvec_regdef *rd) |
| 3363 | { |
| 3364 | unsigned long sz, a, m; |
| 3365 | |
| 3366 | if (unsigned_from_buf(b, &sz)) return (-1); |
| 3367 | if (unsigned_from_buf(b, &a)) return (-1); |
| 3368 | if (unsigned_from_buf(b, &m)) return (-1); |
| 3369 | if (sz > (size_t)-1 || a > (size_t)-1 || m > (size_t)-1) |
| 3370 | { buf_break(b); return (-1); } |
| 3371 | rv->buf.sz = sz; rv->buf.a = a; rv->buf.m = m; |
| 3372 | return (0); |
| 3373 | } |
| 3374 | |
| 3375 | /* --- @parse_buffer@ --- * |
| 3376 | * |
| 3377 | * Arguments: @union tvec_regval *rv@ = register value |
| 3378 | * @const struct tvec_regdef *rd@ = register definition |
| 3379 | * @struct tvec_state *tv@ = test-vector state |
| 3380 | * |
| 3381 | * Returns: Zero on success, %$-1$% on error. |
| 3382 | * |
| 3383 | * Use: Parse a register value from an input file. |
| 3384 | * |
| 3385 | * The input format for a buffer value consists of an unsigned |
| 3386 | * integer followed by an optional unit specifier consisting of |
| 3387 | * an SI unit prefix and (optionally) the letter `B'. Unit |
| 3388 | * prefixes denote %%\emph{binary}%% multipliers, not decimal. |
| 3389 | * |
| 3390 | * The buffer is allocated and filled with a distinctive |
| 3391 | * pattern. |
| 3392 | */ |
| 3393 | |
| 3394 | static int parse_buffer(union tvec_regval *rv, |
| 3395 | const struct tvec_regdef *rd, |
| 3396 | struct tvec_state *tv) |
| 3397 | { |
| 3398 | unsigned long sz, a = 0, m = 0; |
| 3399 | int ch, rc; |
| 3400 | |
| 3401 | if (parse_size(tv, &sz, "@;", "buffer length")) { rc = -1; goto end; } |
| 3402 | if (check_string_length(sz, rd->arg.p, tv)) { rc = -1; goto end; } |
| 3403 | |
| 3404 | tvec_skipspc(tv); |
| 3405 | ch = getc(tv->fp); |
| 3406 | if (ch == ';' || ch == '\n') { ungetc(ch, tv->fp); goto done; } |
| 3407 | else if (ch != '@') { rc = tvec_syntax(tv, ch, "`@'"); goto end; } |
| 3408 | |
| 3409 | if (parse_size(tv, &m, "+;", "alignment quantum")) { rc = -1; goto end; } |
| 3410 | if (m == 1) m = 0; |
| 3411 | |
| 3412 | tvec_skipspc(tv); |
| 3413 | ch = getc(tv->fp); |
| 3414 | if (ch == ';' || ch == '\n') { ungetc(ch, tv->fp); goto done; } |
| 3415 | else if (ch != '+') { rc = tvec_syntax(tv, ch, "`+'"); goto end; } |
| 3416 | |
| 3417 | if (parse_size(tv, &a, ";", "alignment offset")) { rc = -1; goto end; } |
| 3418 | if (a >= m) { |
| 3419 | rc = tvec_error(tv, "alignment offset %lu >= quantum %lu", |
| 3420 | (unsigned long)a, (unsigned long)m); |
| 3421 | goto end; |
| 3422 | } |
| 3423 | |
| 3424 | done: |
| 3425 | if (tvec_flushtoeol(tv, 0)) { rc = -1; goto end; } |
| 3426 | rv->buf.sz = sz; rv->buf.a = a; rv->buf.m = m; |
| 3427 | rc = 0; |
| 3428 | end: |
| 3429 | return (rc); |
| 3430 | } |
| 3431 | |
| 3432 | /* --- @dump_buffer@ --- * |
| 3433 | * |
| 3434 | * Arguments: @const union tvec_regval *rv@ = register value |
| 3435 | * @const struct tvec_regdef *rd@ = register definition |
| 3436 | * @unsigned style@ = output style (@TVSF_...@) |
| 3437 | * @const struct gprintf_ops *gops@, @void *gp@ = format output |
| 3438 | * |
| 3439 | * Returns: --- |
| 3440 | * |
| 3441 | * Use: Dump a register value to the format output. |
| 3442 | * |
| 3443 | * Buffer values are dumped as their size with an appropriate |
| 3444 | * unit specifier. A unit prefix is only used if the size is an |
| 3445 | * exact multiple of the relevant power of two. |
| 3446 | */ |
| 3447 | |
| 3448 | static void dump_buffer(const union tvec_regval *rv, |
| 3449 | const struct tvec_regdef *rd, |
| 3450 | unsigned style, |
| 3451 | const struct gprintf_ops *gops, void *go) |
| 3452 | { |
| 3453 | format_size(gops, go, rv->buf.sz, style); |
| 3454 | if (rv->buf.m) { |
| 3455 | gprintf(gops, go, style&TVSF_COMPACT ? "@" : " @ "); |
| 3456 | format_size(gops, go, rv->buf.m, style); |
| 3457 | if (rv->buf.a) { |
| 3458 | gprintf(gops, go, style&TVSF_COMPACT ? "+" : " + "); |
| 3459 | format_size(gops, go, rv->buf.a, style); |
| 3460 | } |
| 3461 | } |
| 3462 | if (!(style&TVSF_COMPACT)) { |
| 3463 | gprintf(gops, go, " ; = %lu", rv->buf.sz); |
| 3464 | if (rv->buf.m) { |
| 3465 | gprintf(gops, go, " @ %lu", rv->buf.m); |
| 3466 | if (rv->buf.a) gprintf(gops, go, " + %lu", rv->buf.a); |
| 3467 | } |
| 3468 | gprintf(gops, go, " = "); format_unsigned_hex(gops, go, rv->buf.sz); |
| 3469 | if (rv->buf.m) { |
| 3470 | gprintf(gops, go, " @ "); format_unsigned_hex(gops, go, rv->buf.m); |
| 3471 | if (rv->buf.a) { |
| 3472 | gprintf(gops, go, " + "); |
| 3473 | format_unsigned_hex(gops, go, rv->buf.a); |
| 3474 | } |
| 3475 | } |
| 3476 | } |
| 3477 | } |
| 3478 | |
| 3479 | /* Buffer type definition. */ |
| 3480 | const struct tvec_regty tvty_buffer = { |
| 3481 | init_buffer, release_buffer, eq_buffer, |
| 3482 | tobuf_buffer, frombuf_buffer, |
| 3483 | parse_buffer, dump_buffer |
| 3484 | }; |
| 3485 | |
| 3486 | /* --- @tvec_initbuffer@ --- * |
| 3487 | * |
| 3488 | * Arguments: @union tvec_regval *rv@ = register value |
| 3489 | * @const union tvec_regval *src@ = source buffer |
| 3490 | * @size_t sz@ = size to allocate |
| 3491 | * |
| 3492 | * Returns: --- |
| 3493 | * |
| 3494 | * Use: Initialize the alignment parameters in @rv@ to match @src@, |
| 3495 | * and the size to @sz@. |
| 3496 | */ |
| 3497 | |
| 3498 | void tvec_initbuffer(union tvec_regval *rv, |
| 3499 | const union tvec_regval *src, size_t sz) |
| 3500 | { rv->buf.sz = sz; rv->buf.a = src->buf.a; rv->buf.m = src->buf.m; } |
| 3501 | |
| 3502 | /* --- @tvec_allocbuffer@ --- * |
| 3503 | * |
| 3504 | * Arguments: @union tvec_regval *rv@ = register value |
| 3505 | * |
| 3506 | * Returns: --- |
| 3507 | * |
| 3508 | * Use: Allocate @sz@ bytes to the buffer and fill the space with a |
| 3509 | * distinctive pattern. |
| 3510 | */ |
| 3511 | |
| 3512 | void tvec_allocbuffer(union tvec_regval *rv) |
| 3513 | { |
| 3514 | unsigned char *p; size_t n; |
| 3515 | |
| 3516 | if (rv->buf.p) xfree(rv->buf.p - rv->buf.off); |
| 3517 | |
| 3518 | if (rv->buf.m < 2) { |
| 3519 | rv->buf.p = xmalloc(rv->buf.sz); rv->buf.off = 0; |
| 3520 | } else { |
| 3521 | p = xmalloc(rv->buf.sz + rv->buf.m - 1); |
| 3522 | n = (size_t)p%rv->buf.m; |
| 3523 | rv->buf.off = (rv->buf.a - n + rv->buf.m)%rv->buf.m; |
| 3524 | rv->buf.p = p + rv->buf.off; |
| 3525 | } |
| 3526 | memset(rv->buf.p, '?', rv->buf.sz); |
| 3527 | } |
| 3528 | |
| 3529 | /*----- That's all, folks -------------------------------------------------*/ |