Commit | Line | Data |
---|---|---|
b64eb60f MW |
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> | |
e63124bc | 33 | #include <float.h> |
b64eb60f | 34 | #include <limits.h> |
e63124bc | 35 | #include <math.h> |
b64eb60f MW |
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" | |
67b5031e | 45 | #include "maths.h" |
b1a20bee | 46 | |
b64eb60f | 47 | #include "tvec.h" |
b1a20bee MW |
48 | #include "tvec-adhoc.h" |
49 | #include "tvec-types.h" | |
b64eb60f MW |
50 | |
51 | /*----- Preliminary utilities ---------------------------------------------*/ | |
52 | ||
67b5031e MW |
53 | /* --- @trivial_release@ --- * |
54 | * | |
55 | * Arguments: @union tvec_regval *rv@ = a register value | |
56 | * @const struct tvec_regdef@ = the register definition | |
57 | * | |
58 | * Returns: --- | |
59 | * | |
60 | * Use: Does nothing. Used for register values which don't retain | |
61 | * resources. | |
62 | */ | |
3efcfd2d MW |
63 | |
64 | static void trivial_release(union tvec_regval *rv, | |
65 | const struct tvec_regdef *rd) | |
66 | { ; } | |
67 | ||
67b5031e MW |
68 | /*----- Integer utilities -------------------------------------------------*/ |
69 | ||
70 | /* --- @unsigned_to_buf@, @signed_to_buf@ --- * | |
71 | * | |
72 | * Arguments: @buf *b@ = buffer to write on | |
73 | * @unsigned long u@ or @long i@ = integer to write | |
74 | * | |
75 | * Returns: Zero on success, @-1@ on failure. | |
76 | * | |
77 | * Use: Write @i@ to the buffer, in big-endian (two's-complement, it | |
78 | * signed) format. | |
79 | */ | |
80 | ||
81 | static int unsigned_to_buf(buf *b, unsigned long u) | |
82 | { kludge64 k; ASSIGN64(k, u); return (buf_putk64l(b, k)); } | |
83 | ||
b64eb60f MW |
84 | static int signed_to_buf(buf *b, long i) |
85 | { | |
86 | kludge64 k; | |
87 | unsigned long u; | |
88 | ||
89 | u = i; | |
90 | if (i >= 0) ASSIGN64(k, u); | |
91 | else { ASSIGN64(k, ~u); CPL64(k, k); } | |
92 | return (buf_putk64l(b, k)); | |
93 | } | |
94 | ||
67b5031e MW |
95 | /* --- @unsigned_from_buf@, @signed_from_buf@ --- * |
96 | * | |
97 | * Arguments: @buf *b@ = buffer to write on | |
98 | * @unsigned long *u_out@ or @long *i_out@ = where to put the | |
99 | * result | |
100 | * | |
101 | * Returns: Zero on success, @-1@ on failure. | |
102 | * | |
103 | * Use: Read an integer, in big-endian (two's-complement, if signed) | |
104 | * format, from the buffer. | |
105 | */ | |
b64eb60f MW |
106 | |
107 | static int unsigned_from_buf(buf *b, unsigned long *u_out) | |
108 | { | |
109 | kludge64 k, ulmax; | |
110 | ||
111 | ASSIGN64(ulmax, ULONG_MAX); | |
112 | if (buf_getk64l(b, &k)) return (-1); | |
adec5584 | 113 | if (CMP64(k, >, ulmax)) { buf_break(b); return (-1); } |
b64eb60f MW |
114 | *u_out = GET64(unsigned long, k); return (0); |
115 | } | |
116 | ||
67b5031e MW |
117 | /* --- @hex_width@ --- * |
118 | * | |
119 | * Arguments: @unsigned long u@ = an integer | |
120 | * | |
121 | * Returns: A suitable number of digits to use in order to display @u@ in | |
122 | * hex. Currently, we select a power of two sufficient to show | |
123 | * the value, but at least 2. | |
124 | */ | |
125 | ||
b64eb60f MW |
126 | static int hex_width(unsigned long u) |
127 | { | |
128 | int wd; | |
129 | unsigned long t; | |
130 | ||
131 | for (t = u >> 4, wd = 4; t >>= wd, wd *= 2, t; ); | |
132 | return (wd/4); | |
133 | } | |
134 | ||
67b5031e MW |
135 | /* --- @format_unsigned_hex@, @format_signed_hex@ --- * |
136 | * | |
137 | * Arguments: @const struct gprintf_ops *gops@ = print operations | |
138 | * @void *go@ = print destination | |
139 | * @unsigned long u@ or @long i@ = integer to print | |
140 | * | |
141 | * Returns: --- | |
142 | * | |
143 | * Use: Print an unsigned or signed integer in hexadecimal. | |
144 | */ | |
145 | ||
146 | static void format_unsigned_hex(const struct gprintf_ops *gops, void *go, | |
147 | unsigned long u) | |
148 | { gprintf(gops, go, "0x%0*lx", hex_width(u), u); } | |
149 | ||
150 | static void format_signed_hex(const struct gprintf_ops *gops, void *go, | |
151 | long i) | |
152 | { | |
153 | unsigned long u = i >= 0 ? i : -(unsigned long)i; | |
154 | gprintf(gops, go, "%s0x%0*lx", i < 0 ? "-" : "", hex_width(u), u); | |
155 | } | |
156 | ||
157 | static int signed_from_buf(buf *b, long *i_out) | |
158 | { | |
159 | kludge64 k, lmax, not_lmin; | |
160 | ||
161 | ASSIGN64(lmax, LONG_MAX); ASSIGN64(not_lmin, ~(unsigned long)LONG_MIN); | |
162 | if (buf_getk64l(b, &k)) return (-1); | |
163 | if (CMP64(k, <=, lmax)) *i_out = (long)GET64(unsigned long, k); | |
164 | else { | |
165 | CPL64(k, k); | |
166 | if (CMP64(k, <=, not_lmin)) *i_out = -(long)GET64(unsigned long, k) - 1; | |
adec5584 | 167 | else { buf_break(b); return (-1); } |
67b5031e MW |
168 | } |
169 | return (0); | |
170 | } | |
171 | ||
b1a20bee | 172 | /* --- @check_signed_range@, @check_unsigned_range@ --- * |
67b5031e | 173 | * |
b1a20bee MW |
174 | * Arguments: @long i@ or @unsigned long u@ = an integer |
175 | * @const struct tvec_irange *ir@ or | |
176 | * @const struct tvec_urange *ur@ = range specification, | |
67b5031e MW |
177 | * or null |
178 | * @struct tvec_state *tv@ = test vector state | |
c4ccbbf9 | 179 | * @const char *what@ = description of value |
67b5031e MW |
180 | * |
181 | * Returns: Zero on success, or @-1@ on error. | |
182 | * | |
183 | * Use: Check that the integer is within bounds. If not, report a | |
184 | * suitable error and return a failure indication. | |
185 | */ | |
186 | ||
882a39c1 MW |
187 | static int check_signed_range(long i, |
188 | const struct tvec_irange *ir, | |
c4ccbbf9 | 189 | struct tvec_state *tv, const char *what) |
b64eb60f | 190 | { |
b1a20bee MW |
191 | long ii, aa, m; |
192 | ||
193 | if (ir) { | |
194 | if (ir->min > i || i > ir->max) { | |
195 | tvec_error(tv, "%s %ld out of range (must be in [%ld .. %ld])", | |
196 | what, i, ir->min, ir->max); | |
197 | return (-1); | |
198 | } | |
199 | m = ir->m; if (m > 0) m = -m; | |
200 | if (m && m != -1) { | |
201 | /* Reduce both the integer and the intended residue to the canonical | |
202 | * interval [0, m). This is more awkward than it should be because C | |
203 | * (following CPU designs) adopted an unhelpful definition of integer | |
204 | * division when the dividend is negative. | |
205 | * | |
206 | * Note that I've canonicalized the divisor to be %%\emph{negative}%%, | |
207 | * because in two's-complement arithmetic, the absolute value of the | |
208 | * most negative representable value is not itself representable. The | |
209 | * residue modulo the most negative value will itself be representable. | |
210 | */ | |
211 | ||
212 | ii = i%m; if (ii < 0) ii -= m; | |
213 | aa = ir->a%m; if (aa < 0) aa -= m; | |
214 | if (ii != aa) { | |
215 | tvec_error(tv, "%s %ld == %ld =/= %ld (mod %ld)", | |
216 | what, i, ii, ir->a, ir->m); | |
217 | return (-1); | |
218 | } | |
219 | } | |
882a39c1 MW |
220 | } |
221 | return (0); | |
b64eb60f MW |
222 | } |
223 | ||
882a39c1 MW |
224 | static int check_unsigned_range(unsigned long u, |
225 | const struct tvec_urange *ur, | |
c4ccbbf9 | 226 | struct tvec_state *tv, const char *what) |
b64eb60f | 227 | { |
b1a20bee MW |
228 | unsigned long uu; |
229 | ||
230 | if (ur) { | |
231 | if (ur->min > u || u > ur->max) { | |
232 | tvec_error(tv, "%s %lu out of range (must be in [%lu .. %lu])", | |
233 | what, u, ur->min, ur->max); | |
234 | return (-1); | |
235 | } | |
236 | if (ur->m && ur->m != 1) { | |
237 | uu = u%ur->m; | |
238 | if (uu != ur->a%ur->m) { | |
239 | tvec_error(tv, "%s %lu == %lu =/= %lu (mod %lu)", | |
240 | what, u, uu, ur->a, ur->m); | |
241 | return (-1); | |
242 | } | |
243 | } | |
882a39c1 MW |
244 | } |
245 | return (0); | |
b64eb60f MW |
246 | } |
247 | ||
67b5031e MW |
248 | /* --- @chtodig@ --- * |
249 | * | |
250 | * Arguments: @int ch@ = a character | |
251 | * | |
252 | * Returns: The numeric value of the character as a digit, or @-1@ if | |
253 | * it's not a digit. Letters count as extended digits starting | |
254 | * with value 10; case is not significant. | |
255 | */ | |
256 | ||
3efcfd2d MW |
257 | static int chtodig(int ch) |
258 | { | |
259 | if ('0' <= ch && ch <= '9') return (ch - '0'); | |
260 | else if ('a' <= ch && ch <= 'z') return (ch - 'a' + 10); | |
261 | else if ('A' <= ch && ch <= 'Z') return (ch - 'A' + 10); | |
262 | else return (-1); | |
263 | } | |
264 | ||
67b5031e MW |
265 | /* --- @parse_unsigned_integer@, @parse_signed_integer@ --- * |
266 | * | |
267 | * Arguments: @unsigned long *u_out@, @long *i_out@ = where to put the | |
268 | * result | |
269 | * @const char **q_out@ = where to put the end position | |
270 | * @const char *p@ = pointer to the string to parse | |
271 | * | |
272 | * Returns: Zero on success, @-1@ on error. | |
273 | * | |
274 | * Use: Parse an integer from a string in the test-vector format. | |
275 | * This is mostly extension of the traditional C @strtoul@ | |
276 | * format: supported inputs include: | |
277 | * | |
278 | * * NNN -- a decimal number (even if it starts with `0'); | |
279 | * * 0xNNN -- hexadecimal; | |
280 | * * 0oNNN -- octal; | |
281 | * * 0bNNN -- binary; | |
282 | * * NNrNNN -- base NN. | |
283 | * | |
284 | * Furthermore, single underscores are permitted internally as | |
285 | * an insignificant digit separator. | |
286 | */ | |
287 | ||
3efcfd2d MW |
288 | static int parse_unsigned_integer(unsigned long *u_out, const char **q_out, |
289 | const char *p) | |
290 | { | |
291 | unsigned long u; | |
292 | int ch, d, r; | |
293 | const char *q; | |
294 | unsigned f = 0; | |
67b5031e MW |
295 | #define f_implicit 1u /* implicitly reading base 10 */ |
296 | #define f_digit 2u /* read a real digit */ | |
297 | #define f_uscore 4u /* found an underscore */ | |
298 | ||
299 | /* Initial setup | |
300 | * | |
301 | * This will deal with the traditional `0[box]...' prefixes. We'll leave | |
302 | * our new `NNr...' syntax for later. | |
303 | */ | |
3efcfd2d MW |
304 | if (p[0] != '0' || !p[1]) { |
305 | d = chtodig(*p); if (0 > d || d >= 10) return (-1); | |
306 | r = 10; u = d; p++; f |= f_implicit | f_digit; | |
307 | } else { | |
308 | u = 0; d = chtodig(p[2]); | |
309 | if (d < 0) { r = 10; f |= f_implicit | f_digit; p++; } | |
310 | else if ((p[1] == 'x' || p[1] == 'X') && d < 16) { r = 16; p += 2; } | |
311 | else if ((p[1] == 'o' || p[1] == 'O') && d < 8) { r = 8; p += 2; } | |
312 | else if ((p[1] == 'b' || p[1] == 'B') && d < 2) { r = 2; p += 2; } | |
313 | else { r = 10; f |= f_digit; p++; } | |
314 | } | |
315 | ||
316 | q = p; | |
317 | for (;;) { | |
67b5031e MW |
318 | /* Work through the string a character at a time. */ |
319 | ||
320 | ch = *p; switch (ch) { | |
321 | ||
322 | case '_': | |
323 | /* An underscore is OK if we haven't just seen one. */ | |
324 | ||
325 | if (f&f_uscore) goto done; | |
326 | p++; f = (f&~f_implicit) | f_uscore; | |
327 | break; | |
328 | ||
329 | case 'r': case 'R': | |
330 | /* An `r' is OK if the number so far is small enough to be a sensible | |
331 | * base, and we're scanning decimal implicitly. | |
332 | */ | |
333 | ||
334 | if (!(f&f_implicit) || !u || u >= 36) goto done; | |
335 | d = chtodig(p[1]); if (0 > d || d >= u) goto done; | |
336 | r = u; u = d; f = (f&~f_implicit) | f_digit; p += 2; q = p; | |
337 | break; | |
338 | ||
339 | default: | |
340 | /* Otherwise we expect a valid digit and accumulate it. */ | |
341 | d = chtodig(ch); if (d < 0 || d >= r) goto done; | |
342 | if (u > ULONG_MAX/r) return (-1); | |
343 | u *= r; if (u > ULONG_MAX - d) return (-1); | |
344 | u += d; f = (f&~f_uscore) | f_digit; p++; q = p; | |
345 | break; | |
3efcfd2d MW |
346 | } |
347 | } | |
348 | ||
67b5031e | 349 | done: |
3efcfd2d MW |
350 | if (!(f&f_digit)) return (-1); |
351 | *u_out = u; *q_out = q; return (0); | |
352 | ||
353 | #undef f_implicit | |
354 | #undef f_digit | |
355 | #undef f_uscore | |
356 | } | |
357 | ||
358 | static int parse_signed_integer(long *i_out, const char **q_out, | |
359 | const char *p) | |
360 | { | |
361 | unsigned long u; | |
362 | unsigned f = 0; | |
363 | #define f_neg 1u | |
364 | ||
67b5031e | 365 | /* Read an initial sign. */ |
3efcfd2d MW |
366 | if (*p == '+') p++; |
367 | else if (*p == '-') { f |= f_neg; p++; } | |
368 | ||
67b5031e | 369 | /* Scan an unsigned number. */ |
3efcfd2d MW |
370 | if (parse_unsigned_integer(&u, q_out, p)) return (-1); |
371 | ||
67b5031e | 372 | /* Check for signed overflow and apply the sign. */ |
3efcfd2d MW |
373 | if (!(f&f_neg)) { |
374 | if (u > LONG_MAX) return (-1); | |
375 | *i_out = u; | |
376 | } else { | |
377 | if (u && u - 1 > -(LONG_MIN + 1)) return (-1); | |
378 | *i_out = u ? -(long)(u - 1) - 1 : 0; | |
379 | } | |
380 | ||
381 | return (0); | |
382 | ||
383 | #undef f_neg | |
384 | } | |
385 | ||
67b5031e MW |
386 | /* --- @parse_unsigned@, @parse_signed@ --- * |
387 | * | |
388 | * Arguments: @unsigned long *u_out@ or @long *i_out@ = where to put the | |
389 | * result | |
390 | * @const char *p@ = string to parse | |
391 | * @const struct tvec_urange *ur@ or | |
392 | * @const struct tvec_irange *ir@ = range specification, | |
393 | * or null | |
394 | * @struct tvec_state *tv@ = test vector state | |
395 | * | |
396 | * Returns: Zero on success, @-1@ on error. | |
397 | * | |
398 | * Use: Parse and range-check an integer. Unlike @parse_(un)signed_ | |
399 | * integer@, these functions check that there's no cruft | |
400 | * following the final digit, and report errors as they find | |
401 | * them rather than leaving that to the caller. | |
402 | */ | |
403 | ||
404 | static int parse_unsigned(unsigned long *u_out, const char *p, | |
405 | const struct tvec_urange *ur, | |
406 | struct tvec_state *tv) | |
407 | { | |
408 | unsigned long u; | |
409 | const char *q; | |
410 | ||
411 | if (parse_unsigned_integer(&u, &q, p)) | |
412 | return (tvec_error(tv, "invalid unsigned integer `%s'", p)); | |
413 | if (*q) return (tvec_syntax(tv, *q, "end-of-line")); | |
c4ccbbf9 | 414 | if (check_unsigned_range(u, ur, tv, "integer")) return (-1); |
67b5031e MW |
415 | *u_out = u; return (0); |
416 | } | |
417 | ||
882a39c1 MW |
418 | static int parse_signed(long *i_out, const char *p, |
419 | const struct tvec_irange *ir, | |
420 | struct tvec_state *tv) | |
b64eb60f | 421 | { |
b64eb60f | 422 | long i; |
3efcfd2d | 423 | const char *q; |
b64eb60f | 424 | |
3efcfd2d MW |
425 | if (parse_signed_integer(&i, &q, p)) |
426 | return (tvec_error(tv, "invalid signed integer `%s'", p)); | |
427 | if (*q) return (tvec_syntax(tv, *q, "end-of-line")); | |
c4ccbbf9 | 428 | if (check_signed_range(i, ir, tv, "integer")) return (-1); |
3efcfd2d | 429 | *i_out = i; return (0); |
b64eb60f | 430 | } |
adec5584 MW |
431 | static const char size_units[] = "kMGTPEZY"; |
432 | ||
c4ccbbf9 | 433 | /* --- @parse_szint@ --- * |
adec5584 MW |
434 | * |
435 | * Arguments: @struct tvec_state *tv@ = test-vector state | |
c4ccbbf9 | 436 | * @unsigned long *u_out@ = where to put the answer |
adec5584 MW |
437 | * @const char *delims@ = delimiters |
438 | * @const char *what@ = description of what we're parsing | |
439 | * | |
440 | * Returns: Zero on success, %$-1$% on failure. | |
441 | * | |
442 | * Use: Parse a memory size. | |
443 | */ | |
444 | ||
c4ccbbf9 MW |
445 | static int parse_szint(struct tvec_state *tv, unsigned long *u_out, |
446 | const char *delims, const char *what) | |
adec5584 MW |
447 | { |
448 | dstr d = DSTR_INIT; | |
449 | const char *p, *unit; | |
450 | unsigned long u, t; | |
451 | int rc; | |
452 | unsigned f = 0; | |
453 | #define f_range 1u | |
454 | ||
455 | if (tvec_readword(tv, &d, 0, delims, what)) { rc = -1; goto end; } | |
456 | p = d.buf; | |
457 | if (parse_unsigned_integer(&u, &p, p)) goto bad; | |
458 | if (!*p) tvec_readword(tv, &d, &p, delims, 0); | |
459 | ||
adec5584 | 460 | for (t = u, unit = size_units; *unit; unit++) { |
c4ccbbf9 | 461 | if (t > ULONG_MAX/1024) f |= f_range; |
adec5584 MW |
462 | else t *= 1024; |
463 | if (*p == *unit) { | |
464 | if (f&f_range) goto rangerr; | |
465 | u = t; p++; break; | |
466 | } | |
467 | } | |
468 | if (*p == 'B') p++; | |
469 | if (*p) goto bad; | |
470 | ||
471 | *u_out = u; rc = 0; | |
472 | end: | |
473 | dstr_destroy(&d); | |
474 | return (rc); | |
475 | ||
476 | bad: | |
477 | tvec_error(tv, "invalid %s `%s'", what, d.buf); | |
478 | rc = -1; goto end; | |
479 | ||
480 | rangerr: | |
481 | tvec_error(tv, "%s `%s' out of range", what, d.buf); | |
482 | rc = -1; goto end; | |
483 | ||
484 | #undef f_range | |
485 | } | |
486 | ||
487 | /* --- @format_size@ --- * | |
488 | * | |
489 | * Arguments: @const struct gprintf_ops *gops@ = print operations | |
490 | * @void *go@ = print destination | |
491 | * @unsigned long u@ = a size | |
492 | * @unsigned style@ = style (@TVSF_...@) | |
493 | * | |
494 | * Returns: --- | |
495 | * | |
496 | * Use: Format @u@ as a size in bytes to the destination, expressing | |
497 | * it with a unit prefix if this is possible exactly. | |
498 | */ | |
499 | ||
500 | static void format_size(const struct gprintf_ops *gops, void *go, | |
501 | unsigned long u, unsigned style) | |
502 | { | |
503 | const char *unit; | |
504 | ||
5c0f2e08 MW |
505 | if (style&TVSF_RAW) |
506 | gprintf(gops, go, "%lu", u); | |
507 | else if (!u || u%1024) | |
adec5584 MW |
508 | gprintf(gops, go, "%lu%sB", u, style&TVSF_COMPACT ? "" : " "); |
509 | else { | |
510 | for (unit = size_units, u /= 1024; | |
511 | !(u%1024) && unit[1]; | |
512 | u /= 1024, unit++); | |
513 | gprintf(gops, go, "%lu%s%cB", u, style&TVSF_COMPACT ? "" : " ", *unit); | |
514 | } | |
515 | } | |
b64eb60f | 516 | |
67b5031e | 517 | /*----- Floating-point utilities ------------------------------------------*/ |
b64eb60f | 518 | |
67b5031e MW |
519 | /* --- @eqish_floating_p@ --- * |
520 | * | |
521 | * Arguments: @double x, y@ = two numbers to compare | |
522 | * @const struct tvec_floatinfo *fi@ = floating-point info | |
523 | * | |
c4ccbbf9 MW |
524 | * Returns: Nonzero if the comparand @x@ is sufficiently close to the |
525 | * reference @y@, or zero if it's definitely different. | |
67b5031e | 526 | */ |
3efcfd2d | 527 | |
67b5031e MW |
528 | static int eqish_floating_p(double x, double y, |
529 | const struct tvec_floatinfo *fi) | |
3efcfd2d | 530 | { |
67b5031e MW |
531 | double t; |
532 | ||
533 | if (NANP(x)) return (NANP(y)); else if (NANP(y)) return (0); | |
534 | if (INFP(x)) return (x == y); else if (INFP(y)) return (0); | |
535 | ||
536 | switch (fi ? fi->f&TVFF_EQMASK : TVFF_EXACT) { | |
537 | case TVFF_EXACT: | |
538 | return (x == y && NEGP(x) == NEGP(y)); | |
539 | case TVFF_ABSDELTA: | |
540 | t = x - y; if (t < 0) t = -t; return (t < fi->delta); | |
541 | case TVFF_RELDELTA: | |
c4ccbbf9 | 542 | t = 1.0 - x/y; if (t < 0) t = -t; return (t < fi->delta); |
67b5031e MW |
543 | default: |
544 | abort(); | |
545 | } | |
b64eb60f MW |
546 | } |
547 | ||
67b5031e MW |
548 | /* --- @format_floating@ --- * |
549 | * | |
550 | * Arguments: @const struct gprintf_ops *gops@ = print operations | |
551 | * @void *go@ = print destination | |
552 | * @double x@ = number to print | |
553 | * | |
554 | * Returns: --- | |
555 | * | |
556 | * Use: Print a floating-point number, accurately. | |
557 | */ | |
3efcfd2d | 558 | |
e63124bc MW |
559 | static void format_floating(const struct gprintf_ops *gops, void *go, |
560 | double x) | |
561 | { | |
562 | int prec; | |
563 | ||
564 | if (NANP(x)) | |
565 | gprintf(gops, go, "#nan"); | |
566 | else if (INFP(x)) | |
567 | gprintf(gops, go, x > 0 ? "#+inf" : "#-inf"); | |
568 | else { | |
569 | /* Ugh. C doesn't provide any function for just printing a | |
570 | * floating-point number /correctly/, i.e., so that you can read the | |
571 | * result back and recover the number you first thought of. There are | |
572 | * complicated algorithms published for doing this, but I really don't | |
573 | * want to get into that here. So we have this. | |
574 | * | |
575 | * The sign doesn't cause significant difficulty so we're going to ignore | |
576 | * it for now. So suppose we're given a number %$x = f b^e$%, in | |
577 | * base-%$b$% format, so %$f b^n$% and %$e$% are integers, with | |
578 | * %$0 \le f < 1$%. We're going to convert it into the nearest integer | |
579 | * of the form %$X = F B^E$%, with similar conditions, only with the | |
580 | * additional requirement that %$X$% is normalized, i.e., that %$X = 0$% | |
581 | * or %$F \ge B^{-N}$%. | |
582 | * | |
583 | * We're rounding to the nearest such %$X$%. If there is to be ambiguity | |
584 | * in the conversion, then some %$x = f b^e$% and the next smallest | |
585 | * representable number %$x' = x + b^{e-n}$% must both map to the same | |
586 | * %$X$%, which means both %$x$% and %$x'$% must be nearer to %$X$% than | |
587 | * any other number representable in the target system. The nest larger | |
588 | * number is %$X' = X + B^{E-N}$%; the next smaller number will normally | |
589 | * be %$W = X - B^{E-N}$%, but if %$F = 1/B$ then the next smaller number | |
590 | * is actually %$X - B^{E-N-1}$%. We ignore this latter possibility in | |
591 | * the pursuit of a conservative estimate (though actually it doesn't | |
592 | * matter). | |
593 | * | |
594 | * If both %$x$% and %$x'$% map to %$X$% then we must have | |
595 | * %$L = X - B^{E-N}/2 \le x$% and %$x + b^{e-n} \le R = X + B^{E-N}/2$%; | |
596 | * so firstly %$f b^e = x \ge L = W + B^{E-N}/2 > W = (F - B^{-N}) B^E$%, | |
597 | * and secondly %$b^{e-n} \le B^{E-N}$%. Since these inequalities are in | |
598 | * opposite senses, we can divide, giving | |
599 | * | |
600 | * %$f b^e/b^{e-n} > (F - B^{-N}) B^E/B^{E-N}$% , | |
601 | * | |
602 | * whence | |
603 | * | |
604 | * %$f b^n > (F - B^{-N}) B^N = F B^N - 1$% . | |
605 | * | |
606 | * Now %$f \le 1 - b^{-n}$%, and %$F \ge B^{-1}$%, so, for this to be | |
607 | * possible, it must be the case that | |
608 | * | |
609 | * %$(1 - b^{-n}) b^n = b^n - 1 > B^{N-1} - 1$% . | |
610 | * | |
611 | * Then rearrange and take logarithms, obtaining | |
612 | * | |
613 | * %$(N - 1) \log B < n \log b$% , | |
614 | * | |
615 | * and so | |
616 | * | |
617 | * %$N < n \log b/\log B + 1$% . | |
618 | * | |
619 | * Recall that this is a necessary condition for a collision to occur; we | |
620 | * are therefore safe whenever | |
621 | * | |
622 | * %$N \ge n \log b/\log B + 1$% ; | |
623 | * | |
624 | * so, taking ceilings, | |
625 | * | |
626 | * %$N \ge \lceil n \log b/\log B \rceil + 1$% . | |
627 | * | |
628 | * So that's why we have this. | |
629 | * | |
b1a20bee MW |
630 | * I'm going to assume that @n = DBL_MANT_DIG@ is sufficiently small |
631 | * that we can calculate this without ending up on the wrong side of an | |
e63124bc MW |
632 | * integer boundary. |
633 | * | |
b1a20bee MW |
634 | * In C11, we have @DBL_DECIMAL_DIG@, which should be the same value |
635 | * only as a constant. Except that modern compilers are more than clever | |
e63124bc MW |
636 | * enough to work out that this is a constant anyway. |
637 | * | |
638 | * This is sometimes an overestimate: we'll print out meaningless digits | |
639 | * that don't represent anything we actually know about the number in | |
640 | * question. To fix that, we'd need a complicated algorithm like Steele | |
641 | * and White's Dragon4, Gay's @dtoa@, or Burger and Dybvig's algorithm | |
642 | * (note that Loitsch's Grisu2 is conservative, and Grisu3 hands off to | |
643 | * something else in difficult situations). | |
644 | */ | |
645 | ||
b1a20bee MW |
646 | #ifdef DBL_DECIMAL_DIG |
647 | prec = DBL_DECIMAL_DIG; | |
648 | #else | |
e63124bc | 649 | prec = ceil(DBL_MANT_DIG*log(FLT_RADIX)/log(10)) + 1; |
b1a20bee | 650 | #endif |
e63124bc MW |
651 | gprintf(gops, go, "%.*g", prec, x); |
652 | } | |
653 | } | |
654 | ||
67b5031e MW |
655 | /* --- @parse_floating@ --- * |
656 | * | |
657 | * Arguments: @double *x_out@ = where to put the result | |
814e42ff | 658 | * @const char *q_out@ = where to leave end pointer, or null |
67b5031e MW |
659 | * @const char *p@ = string to parse |
660 | * @const struct tvec_floatinfo *fi@ = floating-point info | |
661 | * @struct tvec_state *tv@ = test vector state | |
662 | * | |
663 | * Returns: Zero on success, @-1@ on error. | |
664 | * | |
665 | * Use: Parse a floating-point number from a string. Reports any | |
814e42ff | 666 | * necessary errors. If @q_out@ is not null then trailing |
adec5584 MW |
667 | * material is permitted and a pointer to it (or the end of the |
668 | * string) is left in @*q_out@. | |
67b5031e | 669 | */ |
e63124bc | 670 | |
814e42ff | 671 | static int parse_floating(double *x_out, const char **q_out, const char *p, |
e63124bc MW |
672 | const struct tvec_floatinfo *fi, |
673 | struct tvec_state *tv) | |
674 | { | |
675 | const char *pp; char *q; | |
676 | dstr d = DSTR_INIT; | |
677 | double x; | |
678 | int olderr, rc; | |
679 | ||
67b5031e | 680 | /* Check for special tokens. */ |
e63124bc MW |
681 | if (STRCMP(p, ==, "#nan")) { |
682 | #ifdef NAN | |
adec5584 | 683 | if (q_out) *q_out = p + strlen(p); |
e63124bc MW |
684 | x = NAN; rc = 0; |
685 | #else | |
686 | tvec_error(tv, "NaN not supported on this system"); | |
687 | rc = -1; goto end; | |
688 | #endif | |
67b5031e MW |
689 | } |
690 | ||
691 | else if (STRCMP(p, ==, "#inf") || | |
692 | STRCMP(p, ==, "#+inf") || STRCMP(p, ==, "+#inf")) { | |
3efcfd2d | 693 | #ifdef INFINITY |
adec5584 | 694 | if (q_out) *q_out = p + strlen(p); |
e63124bc MW |
695 | x = INFINITY; rc = 0; |
696 | #else | |
697 | tvec_error(tv, "infinity not supported on this system"); | |
698 | rc = -1; goto end; | |
699 | #endif | |
67b5031e MW |
700 | } |
701 | ||
702 | else if (STRCMP(p, ==, "#-inf") || STRCMP(p, ==, "-#inf")) { | |
3efcfd2d | 703 | #ifdef INFINITY |
adec5584 | 704 | if (q_out) *q_out = p + strlen(p); |
e63124bc MW |
705 | x = -INFINITY; rc = 0; |
706 | #else | |
707 | tvec_error(tv, "infinity not supported on this system"); | |
708 | rc = -1; goto end; | |
709 | #endif | |
67b5031e MW |
710 | } |
711 | ||
712 | /* Check that this looks like a number, so we can exclude `strtod' | |
713 | * recognizing its own non-finite number tokens. | |
714 | */ | |
715 | else { | |
e63124bc MW |
716 | pp = p; |
717 | if (*pp == '+' || *pp == '-') pp++; | |
718 | if (*pp == '.') pp++; | |
719 | if (!ISDIGIT(*pp)) { | |
3efcfd2d | 720 | tvec_syntax(tv, *p ? *p : fgetc(tv->fp), "floating-point number"); |
e63124bc MW |
721 | rc = -1; goto end; |
722 | } | |
67b5031e MW |
723 | |
724 | /* Parse the number using the system parser. */ | |
e63124bc | 725 | olderr = errno; errno = 0; |
b1a20bee | 726 | #if __STDC_VERSION__ >= 199901 |
e63124bc | 727 | x = strtod(p, &q); |
b1a20bee MW |
728 | #else |
729 | x = strtold(p, &q); | |
730 | #endif | |
adec5584 MW |
731 | if (q_out) *q_out = q; |
732 | else if (*q) { tvec_syntax(tv, *q, "end-of-line"); rc = -1; goto end; } | |
e63124bc | 733 | if (errno && (errno != ERANGE || (x > 0 ? -x : x) == HUGE_VAL)) { |
814e42ff MW |
734 | tvec_error(tv, "invalid floating-point number `%.*s': %s", |
735 | (int)(q - p), p, strerror(errno)); | |
e63124bc MW |
736 | rc = -1; goto end; |
737 | } | |
738 | errno = olderr; | |
739 | } | |
740 | ||
67b5031e | 741 | /* Check that the number is acceptable. */ |
e63124bc MW |
742 | if (NANP(x) && fi && !(fi->f&TVFF_NANOK)) { |
743 | tvec_error(tv, "#nan not allowed here"); | |
744 | rc = -1; goto end; | |
745 | } | |
67b5031e | 746 | |
b1a20bee MW |
747 | if (fi && |
748 | ((!(fi->f&TVFF_NOMIN) && x < fi->min) || | |
749 | (!(fi->f&TVFF_NOMAX) && x > fi->max)) && | |
750 | !(INFP(x) && (fi->f&(NEGP(x) ? TVFF_NEGINFOK : TVFF_POSINFOK)))) { | |
e63124bc MW |
751 | dstr_puts(&d, "floating-point number "); |
752 | format_floating(&dstr_printops, &d, x); | |
753 | dstr_puts(&d, " out of range (must be in "); | |
754 | if (fi->f&TVFF_NOMIN) | |
755 | dstr_puts(&d, "(#-inf"); | |
756 | else | |
757 | { dstr_putc(&d, '['); format_floating(&dstr_printops, &d, fi->min); } | |
758 | dstr_puts(&d, " .. "); | |
759 | if (fi->f&TVFF_NOMAX) | |
760 | dstr_puts(&d, "#+inf)"); | |
761 | else | |
762 | { format_floating(&dstr_printops, &d, fi->max); dstr_putc(&d, ']'); } | |
763 | dstr_putc(&d, ')'); dstr_putz(&d); | |
764 | tvec_error(tv, "%s", d.buf); rc = -1; goto end; | |
765 | } | |
766 | ||
67b5031e MW |
767 | /* All done. */ |
768 | *x_out = x; rc = 0; | |
769 | end: | |
770 | dstr_destroy(&d); | |
771 | return (rc); | |
772 | } | |
773 | ||
774 | /*----- String utilities --------------------------------------------------*/ | |
775 | ||
776 | /* Special character name table. */ | |
777 | static const struct chartab { | |
778 | const char *name; /* character name */ | |
779 | int ch; /* character value */ | |
780 | unsigned f; /* flags: */ | |
781 | #define CTF_PREFER 1u /* preferred name */ | |
782 | #define CTF_SHORT 2u /* short name (compact style) */ | |
783 | } chartab[] = { | |
784 | { "#eof", EOF, CTF_PREFER | CTF_SHORT }, | |
785 | { "#nul", '\0', CTF_PREFER }, | |
786 | { "#bell", '\a', CTF_PREFER }, | |
787 | { "#ding", '\a', 0 }, | |
788 | { "#bel", '\a', CTF_SHORT }, | |
789 | { "#backspace", '\b', CTF_PREFER }, | |
790 | { "#bs", '\b', CTF_SHORT }, | |
791 | { "#escape", '\x1b', CTF_PREFER }, | |
792 | { "#esc", '\x1b', CTF_SHORT }, | |
793 | { "#formfeed", '\f', CTF_PREFER }, | |
794 | { "#ff", '\f', CTF_SHORT }, | |
795 | { "#newline", '\n', CTF_PREFER }, | |
796 | { "#linefeed", '\n', 0 }, | |
797 | { "#lf", '\n', CTF_SHORT }, | |
798 | { "#nl", '\n', 0 }, | |
799 | { "#return", '\r', CTF_PREFER }, | |
800 | { "#carriage-return", '\r', 0 }, | |
801 | { "#cr", '\r', CTF_SHORT }, | |
802 | { "#tab", '\t', CTF_PREFER | CTF_SHORT }, | |
803 | { "#horizontal-tab", '\t', 0 }, | |
804 | { "#ht", '\t', 0 }, | |
805 | { "#vertical-tab", '\v', CTF_PREFER }, | |
806 | { "#vt", '\v', CTF_SHORT }, | |
807 | { "#space", ' ', 0 }, | |
808 | { "#spc", ' ', CTF_SHORT }, | |
809 | { "#delete", '\x7f', CTF_PREFER }, | |
810 | { "#del", '\x7f', CTF_SHORT }, | |
811 | { 0, 0, 0 } | |
812 | }; | |
813 | ||
814 | /* --- @find_charname@ --- * | |
815 | * | |
816 | * Arguments: @int ch@ = character to match | |
817 | * @unsigned f@ = flags (@CTF_...@) to match | |
818 | * | |
819 | * Returns: The name of the character, or null if no match is found. | |
820 | * | |
821 | * Use: Looks up a name for a character. Specifically, it returns | |
822 | * the first entry in the @chartab@ table which matches @ch@ and | |
823 | * which has one of the flags @f@ set. | |
824 | */ | |
825 | ||
826 | static const char *find_charname(int ch, unsigned f) | |
827 | { | |
828 | const struct chartab *ct; | |
829 | ||
830 | for (ct = chartab; ct->name; ct++) | |
831 | if (ct->ch == ch && (ct->f&f)) return (ct->name); | |
832 | return (0); | |
833 | } | |
834 | ||
835 | /* --- @read_charname@ --- * | |
836 | * | |
837 | * Arguments: @int *ch_out@ = where to put the character | |
838 | * @const char *p@ = character name | |
839 | * @unsigned f@ = flags (@TCF_...@) | |
840 | * | |
841 | * Returns: Zero if a match was found, @-1@ if not. | |
842 | * | |
843 | * Use: Looks up a character by name. If @RCF_EOFOK@ is set in @f@, | |
844 | * then the @EOF@ marker can be matched; otherwise it can't. | |
845 | */ | |
846 | ||
847 | #define RCF_EOFOK 1u | |
848 | static int read_charname(int *ch_out, const char *p, unsigned f) | |
849 | { | |
850 | const struct chartab *ct; | |
851 | ||
852 | for (ct = chartab; ct->name; ct++) | |
853 | if (STRCMP(p, ==, ct->name) && ((f&RCF_EOFOK) || ct->ch >= 0)) | |
854 | { *ch_out = ct->ch; return (0); } | |
855 | return (-1); | |
856 | } | |
857 | ||
858 | /* --- @format_charesc@ --- * | |
859 | * | |
860 | * Arguments: @const struct gprintf_ops *gops@ = print operations | |
861 | * @void *go@ = print destination | |
862 | * @int ch@ = character to format | |
863 | * @unsigned f@ = flags (@FCF_...@) | |
864 | * | |
865 | * Returns: --- | |
866 | * | |
867 | * Use: Format a character as an escape sequence, possibly as part of | |
868 | * a larger string. If @FCF_BRACE@ is set in @f@, then put | |
869 | * braces around a `\x...' code, so that it's suitable for use | |
870 | * in a longer string. | |
871 | */ | |
872 | ||
873 | #define FCF_BRACE 1u | |
874 | static void format_charesc(const struct gprintf_ops *gops, void *go, | |
875 | int ch, unsigned f) | |
876 | { | |
877 | switch (ch) { | |
878 | case '\a': gprintf(gops, go, "\\a"); break; | |
879 | case '\b': gprintf(gops, go, "\\b"); break; | |
880 | case '\x1b': gprintf(gops, go, "\\e"); break; | |
881 | case '\f': gprintf(gops, go, "\\f"); break; | |
882 | case '\r': gprintf(gops, go, "\\r"); break; | |
883 | case '\n': gprintf(gops, go, "\\n"); break; | |
884 | case '\t': gprintf(gops, go, "\\t"); break; | |
885 | case '\v': gprintf(gops, go, "\\v"); break; | |
886 | case '\\': gprintf(gops, go, "\\\\"); break; | |
887 | case '\'': gprintf(gops, go, "\\'"); break; | |
888 | case '\0': | |
889 | if (f&FCF_BRACE) gprintf(gops, go, "\\{0}"); | |
890 | else gprintf(gops, go, "\\0"); | |
891 | break; | |
892 | default: | |
893 | if (f&FCF_BRACE) | |
894 | gprintf(gops, go, "\\x{%0*x}", hex_width(UCHAR_MAX), ch); | |
895 | else | |
896 | gprintf(gops, go, "\\x%0*x", hex_width(UCHAR_MAX), ch); | |
897 | break; | |
898 | } | |
899 | } | |
900 | ||
901 | /* --- @format_char@ --- * | |
902 | * | |
903 | * Arguments: @const struct gprintf_ops *gops@ = print operations | |
904 | * @void *go@ = print destination | |
905 | * @int ch@ = character to format | |
906 | * | |
907 | * Returns: --- | |
908 | * | |
909 | * Use: Format a single character. | |
910 | */ | |
911 | ||
912 | static void format_char(const struct gprintf_ops *gops, void *go, int ch) | |
913 | { | |
914 | switch (ch) { | |
915 | case '\\': case '\'': escape: | |
916 | gprintf(gops, go, "'"); | |
917 | format_charesc(gops, go, ch, 0); | |
918 | gprintf(gops, go, "'"); | |
919 | break; | |
920 | default: | |
921 | if (!isprint(ch)) goto escape; | |
922 | gprintf(gops, go, "'%c'", ch); | |
923 | break; | |
924 | } | |
925 | } | |
926 | ||
b1a20bee MW |
927 | /* --- @fill_pattern@ --- * |
928 | * | |
929 | * Arguments: @void *p@ = destination pointer | |
930 | * @size_t sz@ = destination buffer size | |
931 | * @const void *pat@ = pointer to pattern | |
932 | * @size_t patsz@ = pattern size | |
933 | * | |
934 | * Returns: --- | |
935 | * | |
936 | * Use: Fill the destination buffer with as many copies of the | |
937 | * pattern as will fit, followed by as many initial bytes of the | |
938 | * pattern will fit in the remaining space. | |
939 | */ | |
940 | ||
941 | static void fill_pattern(void *p, size_t sz, const void *pat, size_t patsz) | |
942 | { | |
943 | unsigned char *q = p; | |
944 | ||
945 | if (patsz == 1) | |
946 | memset(q, *(unsigned char *)pat, sz); | |
947 | else { | |
948 | if (sz > patsz) { | |
949 | memcpy(q, pat, patsz); pat = q; q += patsz; sz -= patsz; | |
950 | while (sz > patsz) | |
951 | { memcpy(q, pat, patsz); q += patsz; sz -= patsz; patsz *= 2; } | |
952 | } | |
953 | memcpy(q, pat, sz); | |
954 | } | |
955 | } | |
956 | ||
67b5031e MW |
957 | /* --- @maybe_format_unsigned_char@, @maybe_format_signed_char@ --- * |
958 | * | |
959 | * Arguments: @const struct gprintf_ops *gops@ = print operations | |
960 | * @void *go@ = print destination | |
961 | * @unsigned long u@ or @long i@ = an integer | |
962 | * | |
963 | * Returns: --- | |
964 | * | |
965 | * Use: Format a (signed or unsigned) integer as a character, if it's | |
966 | * in range, printing something like `= 'q''. It's assumed that | |
967 | * a comment marker has already been output. | |
968 | */ | |
969 | ||
970 | static void maybe_format_unsigned_char | |
971 | (const struct gprintf_ops *gops, void *go, unsigned long u) | |
972 | { | |
973 | const char *p; | |
974 | ||
975 | p = find_charname(u, CTF_PREFER); | |
976 | if (p) gprintf(gops, go, " = %s", p); | |
977 | if (u < UCHAR_MAX) | |
978 | { gprintf(gops, go, " = "); format_char(gops, go, u); } | |
e63124bc MW |
979 | } |
980 | ||
67b5031e MW |
981 | static void maybe_format_signed_char |
982 | (const struct gprintf_ops *gops, void *go, long i) | |
b64eb60f | 983 | { |
67b5031e MW |
984 | const char *p; |
985 | ||
986 | p = find_charname(i, CTF_PREFER); | |
987 | if (p) gprintf(gops, go, " = %s", p); | |
988 | if (0 <= i && i < UCHAR_MAX) | |
989 | { gprintf(gops, go, " = "); format_char(gops, go, i); } | |
b64eb60f MW |
990 | } |
991 | ||
67b5031e MW |
992 | /* --- @read_charesc@ --- * |
993 | * | |
994 | * Arguments: @int *ch_out@ = where to put the result | |
995 | * @struct tvec_state *tv@ = test vector state | |
996 | * | |
997 | * Returns: Zero on success, @-1@ on error. | |
998 | * | |
999 | * Use: Parse and convert an escape sequence from @tv@'s input | |
1000 | * stream, assuming that the initial `\' has already been read. | |
1001 | * Reports errors as appropriate. | |
1002 | */ | |
1003 | ||
1004 | static int read_charesc(int *ch_out, struct tvec_state *tv) | |
b64eb60f | 1005 | { |
b64eb60f MW |
1006 | int ch, i, esc; |
1007 | unsigned f = 0; | |
1008 | #define f_brace 1u | |
1009 | ||
e63124bc MW |
1010 | ch = getc(tv->fp); |
1011 | switch (ch) { | |
67b5031e MW |
1012 | |
1013 | /* Things we shouldn't find. */ | |
1014 | case EOF: case '\n': return (tvec_syntax(tv, ch, "string escape")); | |
1015 | ||
1016 | /* Single-character escapes. */ | |
e63124bc MW |
1017 | case '\'': *ch_out = '\''; break; |
1018 | case '\\': *ch_out = '\\'; break; | |
1019 | case '"': *ch_out = '"'; break; | |
1020 | case 'a': *ch_out = '\a'; break; | |
1021 | case 'b': *ch_out = '\b'; break; | |
1022 | case 'e': *ch_out = '\x1b'; break; | |
1023 | case 'f': *ch_out = '\f'; break; | |
1024 | case 'n': *ch_out = '\n'; break; | |
1025 | case 'r': *ch_out = '\r'; break; | |
1026 | case 't': *ch_out = '\t'; break; | |
1027 | case 'v': *ch_out = '\v'; break; | |
1028 | ||
67b5031e | 1029 | /* Hex escapes, with and without braces. */ |
e63124bc MW |
1030 | case 'x': |
1031 | ch = getc(tv->fp); | |
1032 | if (ch == '{') { f |= f_brace; ch = getc(tv->fp); } | |
1033 | else f &= ~f_brace; | |
67b5031e MW |
1034 | esc = chtodig(ch); |
1035 | if (esc < 0 || esc >= 16) return (tvec_syntax(tv, ch, "hex digit")); | |
e63124bc | 1036 | for (;;) { |
67b5031e MW |
1037 | ch = getc(tv->fp); i = chtodig(ch); if (i < 0 || i >= 16) break; |
1038 | esc = 16*esc + i; | |
e63124bc MW |
1039 | if (esc > UCHAR_MAX) |
1040 | return (tvec_error(tv, | |
1041 | "character code %d out of range", esc)); | |
1042 | } | |
1043 | if (!(f&f_brace)) ungetc(ch, tv->fp); | |
1044 | else if (ch != '}') return (tvec_syntax(tv, ch, "`}'")); | |
1045 | *ch_out = esc; | |
1046 | break; | |
1047 | ||
67b5031e MW |
1048 | /* Other things, primarily octal escapes. */ |
1049 | case '{': | |
1050 | f |= f_brace; ch = getc(tv->fp); | |
1051 | /* fall through */ | |
e63124bc MW |
1052 | default: |
1053 | if ('0' <= ch && ch < '8') { | |
1054 | i = 1; esc = ch - '0'; | |
1055 | for (;;) { | |
1056 | ch = getc(tv->fp); | |
1057 | if ('0' > ch || ch >= '8') { ungetc(ch, tv->fp); break; } | |
1058 | esc = 8*esc + ch - '0'; | |
1059 | i++; if (i >= 3) break; | |
1060 | } | |
67b5031e MW |
1061 | if (f&f_brace) { |
1062 | ch = getc(tv->fp); | |
1063 | if (ch != '}') return (tvec_syntax(tv, ch, "`}'")); | |
1064 | } | |
e63124bc MW |
1065 | if (esc > UCHAR_MAX) |
1066 | return (tvec_error(tv, | |
1067 | "character code %d out of range", esc)); | |
67b5031e | 1068 | *ch_out = esc; break; |
e63124bc MW |
1069 | } else |
1070 | return (tvec_syntax(tv, ch, "string escape")); | |
1071 | } | |
1072 | ||
67b5031e | 1073 | /* Done. */ |
e63124bc MW |
1074 | return (0); |
1075 | ||
1076 | #undef f_brace | |
1077 | } | |
1078 | ||
67b5031e MW |
1079 | /* --- @read_quoted_string@ --- * |
1080 | * | |
1081 | * Arguments: @dstr *d@ = string to write to | |
1082 | * @int quote@ = initial quote, `'' or `"' | |
1083 | * @struct tvec_state *tv@ = test vector state | |
1084 | * | |
1085 | * Returns: Zero on success, @-1@ on error. | |
1086 | * | |
1087 | * Use: Read the rest of a quoted string into @d@, reporting errors | |
1088 | * as appropriate. | |
1089 | * | |
1090 | * A single-quoted string is entirely literal. A double-quoted | |
1091 | * string may contain C-like escapes. | |
1092 | */ | |
1093 | ||
e63124bc MW |
1094 | static int read_quoted_string(dstr *d, int quote, struct tvec_state *tv) |
1095 | { | |
1096 | int ch; | |
b64eb60f MW |
1097 | |
1098 | for (;;) { | |
1099 | ch = getc(tv->fp); | |
b64eb60f MW |
1100 | switch (ch) { |
1101 | case EOF: case '\n': | |
e63124bc | 1102 | return (tvec_syntax(tv, ch, "`%c'", quote)); |
b64eb60f MW |
1103 | case '\\': |
1104 | if (quote == '\'') goto ordinary; | |
e63124bc | 1105 | ch = getc(tv->fp); if (ch == '\n') { tv->lno++; break; } |
67b5031e | 1106 | ungetc(ch, tv->fp); if (read_charesc(&ch, tv)) return (-1); |
e63124bc | 1107 | goto ordinary; |
b64eb60f MW |
1108 | default: |
1109 | if (ch == quote) goto end; | |
1110 | ordinary: | |
1111 | DPUTC(d, ch); | |
1112 | break; | |
1113 | } | |
1114 | } | |
1115 | ||
1116 | end: | |
1117 | DPUTZ(d); | |
882a39c1 | 1118 | return (0); |
e63124bc | 1119 | } |
b64eb60f | 1120 | |
67b5031e MW |
1121 | /* --- @collect_bare@ --- * |
1122 | * | |
1123 | * Arguments: @dstr *d@ = string to write to | |
1124 | * @struct tvec_state *tv@ = test vector state | |
1125 | * | |
1126 | * Returns: Zero on success, @-1@ on error. | |
1127 | * | |
1128 | * Use: Read barewords and the whitespace between them. Stop when we | |
1129 | * encounter something which can't start a bareword. | |
1130 | */ | |
b64eb60f MW |
1131 | |
1132 | static int collect_bare(dstr *d, struct tvec_state *tv) | |
1133 | { | |
1134 | size_t pos = d->len; | |
1135 | enum { WORD, SPACE, ESCAPE }; unsigned s = WORD; | |
1136 | int ch, rc; | |
1137 | ||
1138 | for (;;) { | |
1139 | ch = getc(tv->fp); | |
1140 | switch (ch) { | |
1141 | case EOF: | |
882a39c1 MW |
1142 | tvec_syntax(tv, ch, "bareword"); |
1143 | rc = -1; goto end; | |
b64eb60f MW |
1144 | case '\n': |
1145 | if (s == ESCAPE) { tv->lno++; goto addch; } | |
1146 | if (s == WORD) pos = d->len; | |
882a39c1 | 1147 | ungetc(ch, tv->fp); if (tvec_nexttoken(tv)) { rc = -1; goto end; } |
b64eb60f MW |
1148 | DPUTC(d, ' '); s = SPACE; |
1149 | break; | |
67b5031e | 1150 | case '"': case '\'': case '!': case '#': case ')': case '}': case ']': |
882a39c1 | 1151 | if (s == SPACE) { ungetc(ch, tv->fp); goto done; } |
b64eb60f MW |
1152 | goto addch; |
1153 | case '\\': | |
1154 | s = ESCAPE; | |
1155 | break; | |
1156 | default: | |
1157 | if (s != ESCAPE && isspace(ch)) { | |
1158 | if (s == WORD) pos = d->len; | |
1159 | DPUTC(d, ch); s = SPACE; | |
1160 | break; | |
1161 | } | |
1162 | addch: | |
1163 | DPUTC(d, ch); s = WORD; | |
1164 | } | |
1165 | } | |
1166 | ||
1167 | done: | |
1168 | if (s == SPACE) d->len = pos; | |
882a39c1 MW |
1169 | DPUTZ(d); rc = 0; |
1170 | end: | |
1171 | return (rc); | |
b64eb60f MW |
1172 | } |
1173 | ||
67b5031e MW |
1174 | /* --- @set_up_encoding@ --- * |
1175 | * | |
1176 | * Arguments: @const codec_class **ccl_out@ = where to put the class | |
1177 | * @unsigned *f_out@ = where to put the flags | |
1178 | * @unsigned code@ = the coding scheme to use (@TVEC_...@) | |
1179 | * | |
1180 | * Returns: --- | |
1181 | * | |
1182 | * Use: Helper for @read_compound_string@ below. | |
1183 | * | |
1184 | * Return the appropriate codec class and flags for @code@. | |
1185 | * Leaves @*ccl_out@ null if the coding scheme doesn't have a | |
1186 | * backing codec class (e.g., @TVCODE_BARE@). | |
1187 | */ | |
1188 | ||
1189 | enum { TVCODE_BARE, TVCODE_HEX, TVCODE_BASE64, TVCODE_BASE32 }; | |
b64eb60f MW |
1190 | static void set_up_encoding(const codec_class **ccl_out, unsigned *f_out, |
1191 | unsigned code) | |
1192 | { | |
1193 | switch (code) { | |
1194 | case TVCODE_BARE: | |
1195 | *ccl_out = 0; *f_out = 0; | |
1196 | break; | |
1197 | case TVCODE_HEX: | |
1198 | *ccl_out = &hex_class; *f_out = CDCF_IGNCASE; | |
1199 | break; | |
1200 | case TVCODE_BASE32: | |
1201 | *ccl_out = &base32_class; *f_out = CDCF_IGNCASE | CDCF_IGNEQPAD; | |
1202 | break; | |
1203 | case TVCODE_BASE64: | |
1204 | *ccl_out = &base64_class; *f_out = CDCF_IGNEQPAD; | |
1205 | break; | |
1206 | default: | |
1207 | abort(); | |
1208 | } | |
1209 | } | |
1210 | ||
67b5031e MW |
1211 | /* --- @flush_codec@ --- * |
1212 | * | |
1213 | * Arguments: @codec *cdc@ = a codec, or null | |
1214 | * @dstr *d@ = output string | |
1215 | * @struct tvec_state *tv@ = test vector state | |
1216 | * | |
1217 | * Returns: Zero on success, @-1@ on error. | |
1218 | * | |
1219 | * Use: Helper for @read_compound_string@ below. | |
1220 | * | |
1221 | * Flush out any final buffered material from @cdc@, and check | |
1222 | * that it's in a good state. Frees the codec on success. Does | |
1223 | * nothing if @cdc@ is null. | |
1224 | */ | |
1225 | ||
1226 | static int flush_codec(codec *cdc, dstr *d, struct tvec_state *tv) | |
1227 | { | |
1228 | int err; | |
1229 | ||
1230 | if (cdc) { | |
1231 | err = cdc->ops->code(cdc, 0, 0, d); | |
1232 | if (err) | |
1233 | return (tvec_error(tv, "invalid %s sequence end: %s", | |
1234 | cdc->ops->c->name, codec_strerror(err))); | |
1235 | cdc->ops->destroy(cdc); | |
1236 | } | |
1237 | return (0); | |
1238 | } | |
1239 | ||
1240 | /* --- @read_compound_string@ --- * | |
1241 | * | |
1242 | * Arguments: @void **p_inout@ = address of output buffer pointer | |
1243 | * @size_t *sz_inout@ = address of buffer size | |
1244 | * @unsigned code@ = initial interpretation of barewords | |
1245 | * @unsigned f@ = other flags (@RCSF_...@) | |
1246 | * @struct tvec_state *tv@ = test vector state | |
1247 | * | |
1248 | * Returns: Zero on success, @-1@ on error. | |
1249 | * | |
1250 | * Use: Parse a compound string, i.e., a sequence of stringish pieces | |
1251 | * which might be quoted strings, character names, or barewords | |
1252 | * to be decoded accoding to @code@, interspersed with | |
1253 | * additional directives. | |
1254 | * | |
1255 | * If the initial buffer pointer is non-null and sufficiently | |
1256 | * large, then it will be reused; otherwise, it is freed and a | |
1257 | * fresh, sufficiently large buffer is allocated and returned. | |
b1a20bee | 1258 | * This buffer unconditionally uses the standard-library arena. |
67b5031e MW |
1259 | */ |
1260 | ||
1261 | #define RCSF_NESTED 1u | |
882a39c1 | 1262 | static int read_compound_string(void **p_inout, size_t *sz_inout, |
67b5031e MW |
1263 | unsigned code, unsigned f, |
1264 | struct tvec_state *tv) | |
b64eb60f | 1265 | { |
67b5031e | 1266 | const codec_class *ccl; unsigned cdf; |
b64eb60f MW |
1267 | codec *cdc; |
1268 | dstr d = DSTR_INIT, w = DSTR_INIT; | |
1269 | char *p; | |
67b5031e MW |
1270 | const char *q; |
1271 | void *pp = 0; size_t sz; | |
1272 | unsigned long n; | |
882a39c1 | 1273 | int ch, err, rc; |
b64eb60f | 1274 | |
67b5031e MW |
1275 | set_up_encoding(&ccl, &cdf, code); cdc = 0; |
1276 | ||
1277 | if (tvec_nexttoken(tv)) return (tvec_syntax(tv, fgetc(tv->fp), "string")); | |
b64eb60f MW |
1278 | do { |
1279 | ch = getc(tv->fp); | |
67b5031e MW |
1280 | switch (ch) { |
1281 | ||
1282 | case ')': case ']': case '}': | |
1283 | /* Close brackets. Leave these for recursive caller if there is one, | |
1284 | * or just complain. | |
1285 | */ | |
1286 | ||
1287 | if (!(f&RCSF_NESTED)) | |
1288 | { rc = tvec_syntax(tv, ch, "string"); goto end; } | |
1289 | ungetc(ch, tv->fp); goto done; | |
1290 | ||
1291 | case '"': case '\'': | |
1292 | /* Quotes. Read a quoted string. */ | |
1293 | ||
1294 | if (cdc && flush_codec(cdc, &d, tv)) { rc = -1; goto end; } | |
1295 | cdc = 0; | |
1296 | if (read_quoted_string(&d, ch, tv)) { rc = -1; goto end; } | |
1297 | break; | |
1298 | ||
1299 | case '#': | |
1300 | /* A named character. */ | |
1301 | ||
1302 | ungetc(ch, tv->fp); | |
1303 | if (cdc && flush_codec(cdc, &d, tv)) { rc = -1; goto end; } | |
1304 | cdc = 0; | |
adec5584 | 1305 | DRESET(&w); tvec_readword(tv, &w, 0, ";", "character name"); |
5c0f2e08 | 1306 | if (STRCMP(w.buf, ==, "#empty")) break; |
67b5031e MW |
1307 | if (read_charname(&ch, w.buf, RCF_EOFOK)) { |
1308 | rc = tvec_error(tv, "unknown character name `%s'", d.buf); | |
1309 | goto end; | |
1310 | } | |
1311 | DPUTC(&d, ch); break; | |
1312 | ||
1313 | case '!': | |
1314 | /* A magic keyword. */ | |
1315 | ||
1316 | if (cdc && flush_codec(cdc, &d, tv)) { rc = -1; goto end; } | |
1317 | cdc = 0; | |
b64eb60f | 1318 | ungetc(ch, tv->fp); |
adec5584 | 1319 | DRESET(&w); tvec_readword(tv, &w, 0, ";", "`!'-keyword"); |
67b5031e MW |
1320 | |
1321 | /* Change bareword coding system. */ | |
1322 | if (STRCMP(w.buf, ==, "!bare")) | |
1323 | { code = TVCODE_BARE; set_up_encoding(&ccl, &cdf, code); } | |
1324 | else if (STRCMP(w.buf, ==, "!hex")) | |
1325 | { code = TVCODE_HEX; set_up_encoding(&ccl, &cdf, code); } | |
1326 | else if (STRCMP(w.buf, ==, "!base32")) | |
1327 | { code = TVCODE_BASE32; set_up_encoding(&ccl, &cdf, code); } | |
1328 | else if (STRCMP(w.buf, ==, "!base64")) | |
1329 | { code = TVCODE_BASE64; set_up_encoding(&ccl, &cdf, code); } | |
1330 | ||
1331 | /* Repeated substrings. */ | |
1332 | else if (STRCMP(w.buf, ==, "!repeat")) { | |
1333 | if (tvec_nexttoken(tv)) { | |
1334 | rc = tvec_syntax(tv, fgetc(tv->fp), "repeat count"); | |
1335 | goto end; | |
1336 | } | |
1337 | DRESET(&w); | |
adec5584 | 1338 | if (tvec_readword(tv, &w, 0, ";{", "repeat count")) |
67b5031e MW |
1339 | { rc = -1; goto end; } |
1340 | if (parse_unsigned_integer(&n, &q, w.buf)) { | |
1341 | rc = tvec_error(tv, "invalid repeat count `%s'", w.buf); | |
1342 | goto end; | |
1343 | } | |
1344 | if (*q) { rc = tvec_syntax(tv, *q, "`{'"); goto end; } | |
1345 | if (tvec_nexttoken(tv)) | |
1346 | { rc = tvec_syntax(tv, fgetc(tv->fp), "`{'"); goto end; } | |
1347 | ch = getc(tv->fp); if (ch != '{') | |
1348 | { rc = tvec_syntax(tv, ch, "`{'"); goto end; } | |
1349 | sz = 0; | |
1350 | if (read_compound_string(&pp, &sz, code, f | RCSF_NESTED, tv)) | |
1351 | { rc = -1; goto end; } | |
1352 | ch = getc(tv->fp); if (ch != '}') | |
1353 | { rc = tvec_syntax(tv, ch, "`}'"); goto end; } | |
1354 | if (sz) { | |
1355 | if (n > (size_t)-1/sz) | |
1356 | { rc = tvec_error(tv, "repeat size out of range"); goto end; } | |
b1a20bee MW |
1357 | n *= sz; |
1358 | dstr_ensure(&d, n); | |
1359 | fill_pattern(d.buf + d.len, n, pp, sz); d.len += n; | |
67b5031e MW |
1360 | } |
1361 | xfree(pp); pp = 0; | |
1362 | } | |
1363 | ||
1364 | /* Anything else is an error. */ | |
1365 | else { | |
1366 | tvec_error(tv, "unknown string keyword `%s'", w.buf); | |
1367 | rc = -1; goto end; | |
1368 | } | |
b64eb60f | 1369 | break; |
67b5031e | 1370 | |
b64eb60f | 1371 | default: |
67b5031e MW |
1372 | /* A bareword. Process it according to the current coding system. */ |
1373 | ||
1374 | switch (code) { | |
1375 | case TVCODE_BARE: | |
1376 | ungetc(ch, tv->fp); | |
1377 | if (collect_bare(&d, tv)) goto done; | |
1378 | break; | |
1379 | default: | |
1380 | assert(ccl); | |
1381 | ungetc(ch, tv->fp); DRESET(&w); | |
adec5584 MW |
1382 | if (tvec_readword(tv, &w, 0, ";", |
1383 | "%s-encoded fragment", ccl->name)) | |
67b5031e MW |
1384 | { rc = -1; goto end; } |
1385 | if (!cdc) cdc = ccl->decoder(cdf); | |
1386 | err = cdc->ops->code(cdc, w.buf, w.len, &d); | |
1387 | if (err) { | |
1388 | tvec_error(tv, "invalid %s fragment `%s': %s", | |
1389 | ccl->name, w.buf, codec_strerror(err)); | |
1390 | rc = -1; goto end; | |
1391 | } | |
1392 | break; | |
1393 | } | |
1394 | break; | |
b64eb60f MW |
1395 | } |
1396 | } while (!tvec_nexttoken(tv)); | |
1397 | ||
1398 | done: | |
67b5031e MW |
1399 | /* Wrap things up. */ |
1400 | if (cdc && flush_codec(cdc, &d, tv)) { rc = -1; goto end; } | |
1401 | cdc = 0; | |
b64eb60f | 1402 | if (*sz_inout <= d.len) |
b1a20bee | 1403 | { free(*p_inout); *p_inout = x_alloc(&arena_stdlib, d.len + 1); } |
b64eb60f | 1404 | p = *p_inout; memcpy(p, d.buf, d.len); p[d.len] = 0; *sz_inout = d.len; |
882a39c1 | 1405 | rc = 0; |
67b5031e | 1406 | |
882a39c1 | 1407 | end: |
67b5031e MW |
1408 | /* Clean up any debris. */ |
1409 | if (cdc) cdc->ops->destroy(cdc); | |
1410 | if (pp) xfree(pp); | |
b64eb60f | 1411 | dstr_destroy(&d); dstr_destroy(&w); |
882a39c1 | 1412 | return (rc); |
b64eb60f MW |
1413 | } |
1414 | ||
b64eb60f MW |
1415 | /*----- Signed and unsigned integer types ---------------------------------*/ |
1416 | ||
c81c35df MW |
1417 | /* --- @init_int@, @init_uint@ --- * |
1418 | * | |
1419 | * Arguments: @union tvec_regval *rv@ = register value | |
1420 | * @const struct tvec_regdef *rd@ = register definition | |
1421 | * | |
1422 | * Returns: --- | |
1423 | * | |
1424 | * Use: Initialize a register value. | |
1425 | * | |
1426 | * Integer values are initialized to zero. | |
1427 | */ | |
1428 | ||
b64eb60f MW |
1429 | static void init_int(union tvec_regval *rv, const struct tvec_regdef *rd) |
1430 | { rv->i = 0; } | |
1431 | ||
1432 | static void init_uint(union tvec_regval *rv, const struct tvec_regdef *rd) | |
1433 | { rv->u = 0; } | |
1434 | ||
c81c35df MW |
1435 | /* --- @eq_int@, @eq_uint@ --- * |
1436 | * | |
1437 | * Arguments: @const union tvec_regval *rv0, *rv1@ = register values | |
1438 | * @const struct tvec_regdef *rd@ = register definition | |
1439 | * | |
1440 | * Returns: Nonzero if the values are equal, zero if unequal | |
1441 | * | |
1442 | * Use: Compare register values for equality. | |
1443 | */ | |
1444 | ||
b64eb60f MW |
1445 | static int eq_int(const union tvec_regval *rv0, const union tvec_regval *rv1, |
1446 | const struct tvec_regdef *rd) | |
1447 | { return (rv0->i == rv1->i); } | |
1448 | ||
1449 | static int eq_uint(const union tvec_regval *rv0, | |
1450 | const union tvec_regval *rv1, | |
1451 | const struct tvec_regdef *rd) | |
1452 | { return (rv0->u == rv1->u); } | |
1453 | ||
c81c35df MW |
1454 | /* --- @tobuf_int@, @tobuf_uint@ --- * |
1455 | * | |
1456 | * Arguments: @buf *b@ = buffer | |
1457 | * @const union tvec_regval *rv@ = register value | |
1458 | * @const struct tvec_regdef *rd@ = register definition | |
1459 | * | |
1460 | * Returns: Zero on success, %$-1$% on failure. | |
1461 | * | |
1462 | * Use: Serialize a register value to a buffer. | |
1463 | * | |
1464 | * Integer values are serialized as little-endian 64-bit signed | |
1465 | * or unsigned integers. | |
1466 | */ | |
1467 | ||
b64eb60f MW |
1468 | static int tobuf_int(buf *b, const union tvec_regval *rv, |
1469 | const struct tvec_regdef *rd) | |
1470 | { return (signed_to_buf(b, rv->i)); } | |
1471 | ||
1472 | static int tobuf_uint(buf *b, const union tvec_regval *rv, | |
1473 | const struct tvec_regdef *rd) | |
1474 | { return (unsigned_to_buf(b, rv->u)); } | |
1475 | ||
c81c35df MW |
1476 | /* --- @frombuf_int@, @frombuf_uint@ --- * |
1477 | * | |
1478 | * Arguments: @buf *b@ = buffer | |
1479 | * @union tvec_regval *rv@ = register value | |
1480 | * @const struct tvec_regdef *rd@ = register definition | |
1481 | * | |
1482 | * Returns: Zero on success, %$-1$% on failure. | |
1483 | * | |
1484 | * Use: Deserialize a register value from a buffer. | |
1485 | * | |
1486 | * Integer values are serialized as 64-bit signed or unsigned | |
1487 | * integers. | |
1488 | */ | |
1489 | ||
b64eb60f MW |
1490 | static int frombuf_int(buf *b, union tvec_regval *rv, |
1491 | const struct tvec_regdef *rd) | |
882a39c1 | 1492 | { return (signed_from_buf(b, &rv->i)); } |
b64eb60f MW |
1493 | |
1494 | static int frombuf_uint(buf *b, union tvec_regval *rv, | |
1495 | const struct tvec_regdef *rd) | |
1496 | { return (unsigned_from_buf(b, &rv->u)); } | |
1497 | ||
c81c35df MW |
1498 | /* --- @parse_int@, @parse_uint@ --- * |
1499 | * | |
1500 | * Arguments: @union tvec_regval *rv@ = register value | |
1501 | * @const struct tvec_regdef *rd@ = register definition | |
1502 | * @struct tvec_state *tv@ = test-vector state | |
1503 | * | |
1504 | * Returns: Zero on success, %$-1$% on error. | |
1505 | * | |
1506 | * Use: Parse a register value from an input file. | |
1507 | * | |
1508 | * Integers may be input in decimal, hex, binary, or octal, | |
1509 | * following approximately usual conventions. | |
1510 | * | |
1511 | * * Signed integers may be preceded with a `+' or `-' sign. | |
1512 | * | |
1513 | * * Decimal integers are just a sequence of decimal digits | |
1514 | * `0' ... `9'. | |
1515 | * | |
1516 | * * Octal integers are a sequence of digits `0' ... `7', | |
1517 | * preceded by `0o' or `0O'. | |
1518 | * | |
1519 | * * Hexadecimal integers are a sequence of digits `0' | |
1520 | * ... `9', `a' ... `f', or `A' ... `F', preceded by `0x' or | |
1521 | * `0X'. | |
1522 | * | |
1523 | * * Radix-B integers are a sequence of digits `0' ... `9', | |
1524 | * `a' ... `f', or `A' ... `F', each with value less than B, | |
1525 | * preceded by `Br' or `BR', where 0 < B < 36 is expressed | |
1526 | * in decimal without any leading `0' or internal | |
1527 | * underscores `_'. | |
1528 | * | |
1529 | * * A digit sequence may contain internal underscore `_' | |
1530 | * separators, but not before or after all of the digits; | |
1531 | * and two consecutive `_' characters are not permitted. | |
1532 | */ | |
1533 | ||
882a39c1 MW |
1534 | static int parse_int(union tvec_regval *rv, const struct tvec_regdef *rd, |
1535 | struct tvec_state *tv) | |
b64eb60f MW |
1536 | { |
1537 | dstr d = DSTR_INIT; | |
882a39c1 | 1538 | int rc; |
b64eb60f | 1539 | |
adec5584 MW |
1540 | if (tvec_readword(tv, &d, 0, ";", "signed integer")) |
1541 | { rc = -1; goto end; } | |
c81c35df MW |
1542 | if (parse_signed(&rv->i, d.buf, rd->arg.p, tv)) { rc = -1; goto end; } |
1543 | if (tvec_flushtoeol(tv, 0)) { rc = -1; goto end; } | |
882a39c1 MW |
1544 | rc = 0; |
1545 | end: | |
b64eb60f | 1546 | dstr_destroy(&d); |
882a39c1 | 1547 | return (rc); |
b64eb60f MW |
1548 | } |
1549 | ||
882a39c1 MW |
1550 | static int parse_uint(union tvec_regval *rv, const struct tvec_regdef *rd, |
1551 | struct tvec_state *tv) | |
b64eb60f MW |
1552 | { |
1553 | dstr d = DSTR_INIT; | |
882a39c1 | 1554 | int rc; |
b64eb60f | 1555 | |
adec5584 MW |
1556 | if (tvec_readword(tv, &d, 0, ";", "unsigned integer")) |
1557 | { rc = -1; goto end; } | |
c81c35df MW |
1558 | if (parse_unsigned(&rv->u, d.buf, rd->arg.p, tv)) { rc = -1; goto end; } |
1559 | if (tvec_flushtoeol(tv, 0)) { rc = -1; goto end; } | |
882a39c1 MW |
1560 | rc = 0; |
1561 | end: | |
b64eb60f | 1562 | dstr_destroy(&d); |
882a39c1 | 1563 | return (rc); |
b64eb60f MW |
1564 | } |
1565 | ||
c81c35df MW |
1566 | /* --- @dump_int@, @dump_uint@ --- * |
1567 | * | |
1568 | * Arguments: @const union tvec_regval *rv@ = register value | |
1569 | * @const struct tvec_regdef *rd@ = register definition | |
1570 | * @unsigned style@ = output style (@TVSF_...@) | |
1571 | * @const struct gprintf_ops *gops@, @void *gp@ = format output | |
1572 | * | |
1573 | * Returns: --- | |
1574 | * | |
1575 | * Use: Dump a register value to the format output. | |
1576 | * | |
1577 | * Integer values are dumped in decimal and, unless compact | |
1578 | * output is requested, hex, and maybe a character, as a | |
1579 | * comment. | |
1580 | */ | |
1581 | ||
b64eb60f MW |
1582 | static void dump_int(const union tvec_regval *rv, |
1583 | const struct tvec_regdef *rd, | |
e63124bc MW |
1584 | unsigned style, |
1585 | const struct gprintf_ops *gops, void *go) | |
b64eb60f | 1586 | { |
5c0f2e08 | 1587 | if (style&TVSF_RAW) gprintf(gops, go, "int:"); |
e63124bc | 1588 | gprintf(gops, go, "%ld", rv->i); |
5c0f2e08 | 1589 | if (!(style&(TVSF_COMPACT | TVSF_RAW))) { |
3efcfd2d MW |
1590 | gprintf(gops, go, " ; = "); |
1591 | format_signed_hex(gops, go, rv->i); | |
1592 | maybe_format_signed_char(gops, go, rv->i); | |
b64eb60f MW |
1593 | } |
1594 | } | |
1595 | ||
1596 | static void dump_uint(const union tvec_regval *rv, | |
1597 | const struct tvec_regdef *rd, | |
e63124bc MW |
1598 | unsigned style, |
1599 | const struct gprintf_ops *gops, void *go) | |
b64eb60f | 1600 | { |
5c0f2e08 | 1601 | if (style&TVSF_RAW) gprintf(gops, go, "uint:"); |
e63124bc | 1602 | gprintf(gops, go, "%lu", rv->u); |
5c0f2e08 | 1603 | if (!(style&(TVSF_COMPACT | TVSF_RAW))) { |
3efcfd2d MW |
1604 | gprintf(gops, go, " ; = "); |
1605 | format_unsigned_hex(gops, go, rv->u); | |
1606 | maybe_format_unsigned_char(gops, go, rv->u); | |
e63124bc | 1607 | } |
b64eb60f MW |
1608 | } |
1609 | ||
c81c35df | 1610 | /* Integer type definitions. */ |
b64eb60f | 1611 | const struct tvec_regty tvty_int = { |
3efcfd2d | 1612 | init_int, trivial_release, eq_int, |
b64eb60f MW |
1613 | tobuf_int, frombuf_int, |
1614 | parse_int, dump_int | |
1615 | }; | |
c81c35df MW |
1616 | const struct tvec_regty tvty_uint = { |
1617 | init_uint, trivial_release, eq_uint, | |
1618 | tobuf_uint, frombuf_uint, | |
1619 | parse_uint, dump_uint | |
1620 | }; | |
b64eb60f | 1621 | |
c81c35df | 1622 | /* Predefined integer ranges. */ |
b64eb60f | 1623 | const struct tvec_irange |
b1a20bee MW |
1624 | tvrange_schar = { SCHAR_MIN, SCHAR_MAX, 0, 0 }, |
1625 | tvrange_short = { SHRT_MIN, SHRT_MAX, 0, 0 }, | |
1626 | tvrange_int = { INT_MIN, INT_MAX, 0, 0 }, | |
1627 | tvrange_long = { LONG_MIN, LONG_MAX, 0, 0 }, | |
1628 | tvrange_sbyte = { -128, 127, 0, 0 }, | |
1629 | tvrange_i16 = { -32768, +32767, 0, 0 }, | |
1630 | tvrange_i32 = { -2147483648, 2147483647, 0, 0 }; | |
b64eb60f | 1631 | const struct tvec_urange |
b1a20bee MW |
1632 | tvrange_uchar = { 0, UCHAR_MAX, 0, 0 }, |
1633 | tvrange_ushort = { 0, USHRT_MAX, 0, 0 }, | |
1634 | tvrange_uint = { 0, UINT_MAX, 0, 0 }, | |
1635 | tvrange_ulong = { 0, ULONG_MAX, 0, 0 }, | |
1636 | tvrange_size = { 0, (size_t)-1, 0, 0 }, | |
1637 | tvrange_byte = { 0, 255, 0, 0 }, | |
1638 | tvrange_u16 = { 0, 65535, 0, 0 }, | |
1639 | tvrange_u32 = { 0, 4294967295, 0, 0 }; | |
b64eb60f | 1640 | |
67b5031e MW |
1641 | /* --- @tvec_claimeq_int@ --- * |
1642 | * | |
1643 | * Arguments: @struct tvec_state *tv@ = test-vector state | |
1644 | * @long i0, i1@ = two signed integers | |
1645 | * @const char *file@, @unsigned @lno@ = calling file and line | |
1646 | * @const char *expr@ = the expression to quote on failure | |
1647 | * | |
1648 | * Returns: Nonzero if @i0@ and @i1@ are equal, otherwise zero. | |
1649 | * | |
1650 | * Use: Check that values of @i0@ and @i1@ are equal. As for | |
1651 | * @tvec_claim@ above, a test case is automatically begun and | |
1652 | * ended if none is already underway. If the values are | |
1653 | * unequal, then @tvec_fail@ is called, quoting @expr@, and the | |
1654 | * mismatched values are dumped: @i0@ is printed as the output | |
1655 | * value and @i1@ is printed as the input reference. | |
1656 | */ | |
1657 | ||
b64eb60f MW |
1658 | int tvec_claimeq_int(struct tvec_state *tv, long i0, long i1, |
1659 | const char *file, unsigned lno, const char *expr) | |
1660 | { | |
3efcfd2d | 1661 | tv->out[0].v.i = i0; tv->in[0].v.i = i1; |
b64eb60f MW |
1662 | return (tvec_claimeq(tv, &tvty_int, 0, file, lno, expr)); |
1663 | } | |
1664 | ||
67b5031e MW |
1665 | /* --- @tvec_claimeq_uint@ --- * |
1666 | * | |
1667 | * Arguments: @struct tvec_state *tv@ = test-vector state | |
1668 | * @unsigned long u0, u1@ = two unsigned integers | |
1669 | * @const char *file@, @unsigned @lno@ = calling file and line | |
1670 | * @const char *expr@ = the expression to quote on failure | |
1671 | * | |
1672 | * Returns: Nonzero if @u0@ and @u1@ are equal, otherwise zero. | |
1673 | * | |
1674 | * Use: Check that values of @u0@ and @u1@ are equal. As for | |
1675 | * @tvec_claim@ above, a test case is automatically begun and | |
1676 | * ended if none is already underway. If the values are | |
1677 | * unequal, then @tvec_fail@ is called, quoting @expr@, and the | |
1678 | * mismatched values are dumped: @u0@ is printed as the output | |
1679 | * value and @u1@ is printed as the input reference. | |
1680 | */ | |
1681 | ||
b64eb60f MW |
1682 | int tvec_claimeq_uint(struct tvec_state *tv, |
1683 | unsigned long u0, unsigned long u1, | |
1684 | const char *file, unsigned lno, const char *expr) | |
1685 | { | |
3efcfd2d | 1686 | tv->out[0].v.u = u0; tv->in[0].v.u = u1; |
b64eb60f MW |
1687 | return (tvec_claimeq(tv, &tvty_uint, 0, file, lno, expr)); |
1688 | } | |
1689 | ||
c4ccbbf9 MW |
1690 | /*----- Size type ---------------------------------------------------------*/ |
1691 | ||
1692 | /* --- @parse_size@ --- * | |
1693 | * | |
1694 | * Arguments: @union tvec_regval *rv@ = register value | |
1695 | * @const struct tvec_regdef *rd@ = register definition | |
1696 | * @struct tvec_state *tv@ = test-vector state | |
1697 | * | |
1698 | * Returns: Zero on success, %$-1$% on error. | |
1699 | * | |
1700 | * Use: Parse a register value from an input file. | |
1701 | * | |
1702 | * The input format for a size value consists of an unsigned | |
1703 | * integer followed by an optional unit specifier consisting of | |
1704 | * an SI unit prefix and (optionally) the letter `B'. */ | |
1705 | ||
1706 | static int parse_size(union tvec_regval *rv, const struct tvec_regdef *rd, | |
1707 | struct tvec_state *tv) | |
1708 | { | |
1709 | unsigned long sz; | |
1710 | int rc; | |
1711 | ||
1712 | if (parse_szint(tv, &sz, ";", "size")) { rc = -1; goto end; } | |
1713 | if (check_unsigned_range(sz, rd->arg.p, tv, "size")) { rc = -1; goto end; } | |
1714 | if (tvec_flushtoeol(tv, 0)) { rc = -1; goto end; } | |
1715 | rv->u = sz; rc = 0; | |
1716 | end: | |
1717 | return (rc); | |
1718 | } | |
1719 | ||
1720 | /* --- @dump_size@ --- * | |
1721 | * | |
1722 | * Arguments: @const union tvec_regval *rv@ = register value | |
1723 | * @const struct tvec_regdef *rd@ = register definition | |
1724 | * @unsigned style@ = output style (@TVSF_...@) | |
1725 | * @const struct gprintf_ops *gops@, @void *gp@ = format output | |
1726 | * | |
1727 | * Returns: --- | |
1728 | * | |
1729 | * Use: Dump a register value to the format output. | |
1730 | * | |
1731 | * Size values are dumped with a unit specifier, with a unit | |
1732 | * prefox only if the size is an exact multiple of the relevant | |
1733 | * power of two. Unless compact style is requested, the plain | |
1734 | * decimal and hex representations of the value are also | |
1735 | * printed. | |
1736 | */ | |
1737 | ||
1738 | static void dump_size(const union tvec_regval *rv, | |
1739 | const struct tvec_regdef *rd, | |
1740 | unsigned style, | |
1741 | const struct gprintf_ops *gops, void *go) | |
1742 | { | |
5c0f2e08 | 1743 | if (style&TVSF_RAW) gprintf(gops, go, "size:"); |
c4ccbbf9 | 1744 | format_size(gops, go, rv->u, style); |
5c0f2e08 | 1745 | if (!(style&(TVSF_COMPACT | TVSF_RAW))) { |
c4ccbbf9 MW |
1746 | gprintf(gops, go, " ; = %lu", (unsigned long)rv->u); |
1747 | gprintf(gops, go, " = "); format_unsigned_hex(gops, go, rv->u); | |
1748 | maybe_format_unsigned_char(gops, go, rv->u); | |
1749 | } | |
1750 | } | |
1751 | ||
1752 | /* Size type definitions. */ | |
1753 | const struct tvec_regty tvty_size = { | |
1754 | init_uint, trivial_release, eq_uint, | |
1755 | tobuf_uint, frombuf_uint, | |
1756 | parse_size, dump_size | |
1757 | }; | |
1758 | ||
1759 | /* --- @tvec_claimeq_size@ --- * | |
1760 | * | |
1761 | * Arguments: @struct tvec_state *tv@ = test-vector state | |
1762 | * @unsigned long sz0, sz1@ = two sizes | |
1763 | * @const char *file@, @unsigned @lno@ = calling file and line | |
1764 | * @const char *expr@ = the expression to quote on failure | |
1765 | * | |
1766 | * Returns: Nonzero if @sz0@ and @sz1@ are equal, otherwise zero. | |
1767 | * | |
1768 | * Use: Check that values of @u0@ and @u1@ are equal. As for | |
1769 | * @tvec_claim@ above, a test case is automatically begun and | |
1770 | * ended if none is already underway. If the values are | |
1771 | * unequal, then @tvec_fail@ is called, quoting @expr@, and the | |
1772 | * mismatched values are dumped: @u0@ is printed as the output | |
1773 | * value and @u1@ is printed as the input reference. | |
1774 | */ | |
1775 | ||
1776 | int tvec_claimeq_size(struct tvec_state *tv, | |
1777 | unsigned long sz0, unsigned long sz1, | |
1778 | const char *file, unsigned lno, const char *expr) | |
1779 | { | |
1780 | tv->out[0].v.u = sz0; tv->in[0].v.u = sz1; | |
1781 | return (tvec_claimeq(tv, &tvty_size, 0, file, lno, expr)); | |
1782 | } | |
1783 | ||
3efcfd2d | 1784 | /*----- Floating-point type -----------------------------------------------*/ |
e63124bc | 1785 | |
814e42ff | 1786 | /* --- @int_float@ --- * |
c81c35df MW |
1787 | * |
1788 | * Arguments: @union tvec_regval *rv@ = register value | |
1789 | * @const struct tvec_regdef *rd@ = register definition | |
1790 | * | |
1791 | * Returns: --- | |
1792 | * | |
1793 | * Use: Initialize a register value. | |
1794 | * | |
1795 | * Floating-point values are initialized to zero. | |
1796 | */ | |
1797 | ||
e63124bc MW |
1798 | static void init_float(union tvec_regval *rv, const struct tvec_regdef *rd) |
1799 | { rv->f = 0.0; } | |
e63124bc | 1800 | |
c81c35df MW |
1801 | /* --- @eq_float@ --- * |
1802 | * | |
1803 | * Arguments: @const union tvec_regval *rv0, *rv1@ = register values | |
1804 | * @const struct tvec_regdef *rd@ = register definition | |
1805 | * | |
1806 | * Returns: Nonzero if the values are equal, zero if unequal | |
1807 | * | |
1808 | * Use: Compare register values for equality. | |
1809 | * | |
1810 | * Floating-point values may be considered equal if their | |
1811 | * absolute or relative difference is sufficiently small, as | |
1812 | * described in the register definition. | |
1813 | */ | |
1814 | ||
e63124bc MW |
1815 | static int eq_float(const union tvec_regval *rv0, |
1816 | const union tvec_regval *rv1, | |
1817 | const struct tvec_regdef *rd) | |
1818 | { return (eqish_floating_p(rv0->f, rv1->f, rd->arg.p)); } | |
1819 | ||
c81c35df MW |
1820 | /* --- @tobuf_float@ --- * |
1821 | * | |
1822 | * Arguments: @buf *b@ = buffer | |
1823 | * @const union tvec_regval *rv@ = register value | |
1824 | * @const struct tvec_regdef *rd@ = register definition | |
1825 | * | |
1826 | * Returns: Zero on success, %$-1$% on failure. | |
1827 | * | |
1828 | * Use: Serialize a register value to a buffer. | |
1829 | * | |
1830 | * Floating-point values are serialized as little-endian | |
1831 | * IEEE 754 Binary64. | |
1832 | */ | |
1833 | ||
e63124bc MW |
1834 | static int tobuf_float(buf *b, const union tvec_regval *rv, |
1835 | const struct tvec_regdef *rd) | |
1836 | { return (buf_putf64l(b, rv->f)); } | |
c81c35df MW |
1837 | |
1838 | /* --- @frombuf_float@ --- * | |
1839 | * | |
1840 | * Arguments: @buf *b@ = buffer | |
1841 | * @union tvec_regval *rv@ = register value | |
1842 | * @const struct tvec_regdef *rd@ = register definition | |
1843 | * | |
1844 | * Returns: Zero on success, %$-1$% on failure. | |
1845 | * | |
1846 | * Use: Deserialize a register value from a buffer. | |
1847 | * | |
1848 | * Floating-point values are serialized as little-endian | |
1849 | * IEEE 754 Binary64. | |
1850 | */ | |
1851 | ||
e63124bc | 1852 | static int frombuf_float(buf *b, union tvec_regval *rv, |
b1a20bee MW |
1853 | const struct tvec_regdef *rd) |
1854 | { | |
1855 | double t; | |
1856 | int rc; | |
1857 | ||
1858 | rc = buf_getf64l(b, &t); if (!rc) rv->f = t; | |
1859 | return (rc); | |
1860 | } | |
e63124bc | 1861 | |
c81c35df MW |
1862 | /* --- @parse_float@ --- * |
1863 | * | |
1864 | * Arguments: @union tvec_regval *rv@ = register value | |
1865 | * @const struct tvec_regdef *rd@ = register definition | |
1866 | * @struct tvec_state *tv@ = test-vector state | |
1867 | * | |
1868 | * Returns: Zero on success, %$-1$% on error. | |
1869 | * | |
1870 | * Use: Parse a register value from an input file. | |
1871 | * | |
1872 | * Floating-point values are either NaN (%|#nan|%, if supported | |
1873 | * by the platform); positive or negative infinity (%|#inf|%, | |
1874 | * %|+#inf|%, or %|#+inf|% (preferring the last), and %|-#inf|% | |
1875 | * or %|#-inf|% (preferring the latter), if supported by the | |
1876 | * platform); or a number in strtod(3) syntax. | |
1877 | */ | |
1878 | ||
e63124bc MW |
1879 | static int parse_float(union tvec_regval *rv, const struct tvec_regdef *rd, |
1880 | struct tvec_state *tv) | |
1881 | { | |
1882 | dstr d = DSTR_INIT; | |
1883 | int rc; | |
1884 | ||
adec5584 | 1885 | if (tvec_readword(tv, &d, 0, ";", "floating-point number")) |
e63124bc | 1886 | { rc = -1; goto end; } |
814e42ff MW |
1887 | if (parse_floating(&rv->f, 0, d.buf, rd->arg.p, tv)) |
1888 | { rc = -1; goto end; } | |
c81c35df | 1889 | if (tvec_flushtoeol(tv, 0)) { rc = -1; goto end; } |
e63124bc MW |
1890 | rc = 0; |
1891 | end: | |
1892 | dstr_destroy(&d); | |
1893 | return (rc); | |
1894 | } | |
1895 | ||
c81c35df MW |
1896 | /* --- @dump_float@ --- * |
1897 | * | |
1898 | * Arguments: @const union tvec_regval *rv@ = register value | |
1899 | * @const struct tvec_regdef *rd@ = register definition | |
1900 | * @unsigned style@ = output style (@TVSF_...@) | |
1901 | * @const struct gprintf_ops *gops@, @void *gp@ = format output | |
1902 | * | |
1903 | * Returns: --- | |
1904 | * | |
1905 | * Use: Dump a register value to the format output. | |
1906 | * | |
1907 | * Floating-point values are dumped in decimal or as a special | |
1908 | * token beginning with `%|#|%'. Some effort is taken to ensure | |
1909 | * that the output is sufficient to uniquely identify the | |
1910 | * original value, but, honestly, C makes this really hard. | |
1911 | */ | |
1912 | ||
e63124bc MW |
1913 | static void dump_float(const union tvec_regval *rv, |
1914 | const struct tvec_regdef *rd, | |
1915 | unsigned style, | |
1916 | const struct gprintf_ops *gops, void *go) | |
5c0f2e08 MW |
1917 | { |
1918 | if (style&TVSF_RAW) gprintf(gops, go, "float:"); | |
1919 | format_floating(gops, go, rv->f); | |
1920 | } | |
e63124bc | 1921 | |
c81c35df | 1922 | /* Floating-point type definition. */ |
e63124bc | 1923 | const struct tvec_regty tvty_float = { |
3efcfd2d | 1924 | init_float, trivial_release, eq_float, |
e63124bc MW |
1925 | tobuf_float, frombuf_float, |
1926 | parse_float, dump_float | |
1927 | }; | |
1928 | ||
c81c35df MW |
1929 | /* Predefined floating-point ranges. */ |
1930 | const struct tvec_floatinfo | |
b1a20bee MW |
1931 | tvflt_float = { TVFF_EXACT | TVFF_INFOK | TVFF_NANOK, |
1932 | -FLT_MAX, FLT_MAX, 0.0 }, | |
1933 | tvflt_double = { TVFF_EXACT | TVFF_INFOK | TVFF_NANOK, | |
1934 | -DBL_MAX, DBL_MAX, 0.0 }, | |
c81c35df MW |
1935 | tvflt_finite = { TVFF_EXACT, -DBL_MAX, DBL_MAX, 0.0 }, |
1936 | tvflt_nonneg = { TVFF_EXACT, 0, DBL_MAX, 0.0 }; | |
1937 | ||
67b5031e MW |
1938 | /* --- @tvec_claimeqish_float@ --- * |
1939 | * | |
1940 | * Arguments: @struct tvec_state *tv@ = test-vector state | |
1941 | * @double f0, f1@ = two floating-point numbers | |
1942 | * @unsigned f@ = flags (@TVFF_...@) | |
1943 | * @double delta@ = maximum tolerable difference | |
1944 | * @const char *file@, @unsigned @lno@ = calling file and line | |
1945 | * @const char *expr@ = the expression to quote on failure | |
1946 | * | |
c4ccbbf9 | 1947 | * Returns: Nonzero if @f0@ and @f1@ are sufficiently close, otherwise |
67b5031e MW |
1948 | * zero. |
1949 | * | |
1950 | * Use: Check that values of @f0@ and @f1@ are sufficiently close. | |
1951 | * As for @tvec_claim@ above, a test case is automatically begun | |
1952 | * and ended if none is already underway. If the values are | |
1953 | * too far apart, then @tvec_fail@ is called, quoting @expr@, | |
1954 | * and the mismatched values are dumped: @f0@ is printed as the | |
1955 | * output value and @f1@ is printed as the input reference. | |
1956 | * | |
1957 | * The details for the comparison are as follows. | |
1958 | * | |
1959 | * * A NaN value matches any other NaN, and nothing else. | |
1960 | * | |
1961 | * * An infinity matches another infinity of the same sign, | |
1962 | * and nothing else. | |
1963 | * | |
1964 | * * If @f&TVFF_EQMASK@ is @TVFF_EXACT@, then any | |
1965 | * representable number matches only itself: in particular, | |
1966 | * positive and negative zero are considered distinct. | |
1967 | * (This allows tests to check that they land on the correct | |
1968 | * side of branch cuts, for example.) | |
1969 | * | |
1970 | * * If @f&TVFF_EQMASK@ is @TVFF_ABSDELTA@, then %$x$% matches | |
1971 | * %$y$% when %$|x - y| < \delta$%. | |
1972 | * | |
1973 | * * If @f&TVFF_EQMASK@ is @TVFF_RELDELTA@, then %$x$% matches | |
c4ccbbf9 MW |
1974 | * %$y$% when %$|1 - x/y| < \delta$%. (Note that this |
1975 | * criterion is asymmetric. Write %$x \approx_\delta y$% | |
1976 | * if and only if %$|1 - x/y < \delta$%. Then, for example, | |
1977 | * if %$y/(1 + \delta) < x < y (1 - \delta)$%, then | |
1978 | * %$x \approx_\delta y$%, but %$y \not\approx_\delta x$%.) | |
67b5031e MW |
1979 | */ |
1980 | ||
e63124bc MW |
1981 | int tvec_claimeqish_float(struct tvec_state *tv, |
1982 | double f0, double f1, unsigned f, double delta, | |
1983 | const char *file, unsigned lno, | |
1984 | const char *expr) | |
1985 | { | |
1986 | struct tvec_floatinfo fi; | |
1987 | union tvec_misc arg; | |
1988 | ||
1989 | fi.f = f; fi.min = fi.max = 0.0; fi.delta = delta; arg.p = &fi; | |
3efcfd2d | 1990 | tv->out[0].v.f = f0; tv->in[0].v.f = f1; |
e63124bc MW |
1991 | return (tvec_claimeq(tv, &tvty_float, &arg, file, lno, expr)); |
1992 | } | |
e63124bc | 1993 | |
67b5031e MW |
1994 | /* --- @tvec_claimeq_float@ --- * |
1995 | * | |
1996 | * Arguments: @struct tvec_state *tv@ = test-vector state | |
1997 | * @double f0, f1@ = two floating-point numbers | |
1998 | * @const char *file@, @unsigned @lno@ = calling file and line | |
1999 | * @const char *expr@ = the expression to quote on failure | |
2000 | * | |
c4ccbbf9 | 2001 | * Returns: Nonzero if @f0@ and @f1@ are identical, otherwise zero. |
67b5031e MW |
2002 | * |
2003 | * Use: Check that values of @f0@ and @f1@ are identical. The | |
2004 | * function is exactly equivalent to @tvec_claimeqish_float@ | |
2005 | * with @f == TVFF_EXACT@. | |
2006 | */ | |
2007 | ||
2008 | int tvec_claimeq_float(struct tvec_state *tv, | |
2009 | double f0, double f1, | |
2010 | const char *file, unsigned lno, | |
2011 | const char *expr) | |
2012 | { | |
2013 | return (tvec_claimeqish_float(tv, f0, f1, TVFF_EXACT, 0.0, | |
2014 | file, lno, expr)); | |
2015 | } | |
2016 | ||
814e42ff MW |
2017 | /*----- Durations ---------------------------------------------------------*/ |
2018 | ||
2019 | /* A duration is a floating-point number of seconds. Initialization and | |
2020 | * teardown, equality comparison, and serialization are as for floating-point | |
2021 | * values. | |
2022 | */ | |
2023 | ||
2024 | static const struct duration_unit { | |
2025 | const char *unit; | |
2026 | double scale; | |
2027 | unsigned f; | |
2028 | #define DUF_PREFER 1u | |
2029 | } duration_units[] = { | |
2030 | { "Ys", 1e+24, 0 }, | |
2031 | { "Zs", 1e+21, 0 }, | |
2032 | { "Es", 1e+18, 0 }, | |
2033 | { "Ps", 1e+15, 0 }, | |
2034 | { "Ts", 1e+12, 0 }, | |
2035 | { "Gs", 1e+9, 0 }, | |
2036 | { "Ms", 1e+6, 0 }, | |
2037 | { "ks", 1e+3, 0 }, | |
2038 | { "hs", 1e+2, 0 }, | |
2039 | { "das", 1e+1, 0 }, | |
2040 | ||
2041 | { "yr", 31557600.0, DUF_PREFER }, | |
2042 | { "y", 31557600.0, 0 }, | |
2043 | { "day", 86400.0, DUF_PREFER }, | |
2044 | { "dy", 86400.0, 0 }, | |
2045 | { "d", 86400.0, 0 }, | |
2046 | { "hr", 3600.0, DUF_PREFER }, | |
2047 | { "hour", 3600.0, 0 }, | |
2048 | { "h", 3600.0, 0 }, | |
2049 | { "min", 60.0, DUF_PREFER }, | |
2050 | { "m", 60.0, 0 }, | |
2051 | ||
2052 | { "s", 1.0, DUF_PREFER }, | |
2053 | { "sec", 1.0, 0 }, | |
2054 | ||
2055 | { "ds", 1e-1, 0 }, | |
2056 | { "cs", 1e-2, 0 }, | |
2057 | { "ms", 1e-3, DUF_PREFER }, | |
2058 | { "µs", 1e-6, DUF_PREFER }, | |
2059 | { "ns", 1e-9, DUF_PREFER }, | |
2060 | { "ps", 1e-12, DUF_PREFER }, | |
2061 | { "fs", 1e-15, DUF_PREFER }, | |
2062 | { "as", 1e-18, DUF_PREFER }, | |
2063 | { "zs", 1e-21, DUF_PREFER }, | |
2064 | { "ys", 1e-24, DUF_PREFER }, | |
2065 | ||
2066 | { 0 } | |
2067 | }; | |
2068 | ||
13ee7406 MW |
2069 | /* --- @tvec_parsedurunit@ --- * |
2070 | * | |
2071 | * Arguments: @double *scale_out@ = where to leave the scale | |
2072 | * @const char **p_inout@ = input unit string, updated | |
2073 | * | |
2074 | * Returns: Zero on success, %$-1$% on error. | |
2075 | * | |
2076 | * Use: If @*p_inout@ begins with a unit string followed by the end | |
2077 | * of the string or some non-alphanumeric character, then store | |
2078 | * the corresponding scale factor in @*scale_out@, advance | |
2079 | * @*p_inout@ past the unit string, and return zero. Otherwise, | |
2080 | * return %$-1$%. | |
2081 | */ | |
2082 | ||
2083 | int tvec_parsedurunit(double *scale_out, const char **p_inout) | |
2084 | { | |
2085 | const char *p = *p_inout, *q; | |
2086 | const struct duration_unit *u; | |
2087 | size_t n; | |
2088 | ||
2089 | while (ISSPACE(*p)) p++; | |
2090 | for (q = p; *q && ISALNUM(*q); q++); | |
2091 | n = q - p; if (!n) { *scale_out = 1.0; return (0); } | |
2092 | ||
2093 | for (u = duration_units; u->unit; u++) | |
2094 | if (STRNCMP(p, ==, u->unit, n) && !u->unit[n]) | |
2095 | { *scale_out = u->scale; *p_inout = q; return (0); } | |
2096 | return (-1); | |
2097 | } | |
2098 | ||
814e42ff MW |
2099 | /* --- @parse_duration@ --- * |
2100 | * | |
2101 | * Arguments: @union tvec_regval *rv@ = register value | |
2102 | * @const struct tvec_regdef *rd@ = register definition | |
2103 | * @struct tvec_state *tv@ = test-vector state | |
2104 | * | |
2105 | * Returns: Zero on success, %$-1$% on error. | |
2106 | * | |
2107 | * Use: Parse a register value from an input file. | |
2108 | * | |
2109 | * Duration values are finite nonnegative floating-point | |
2110 | * numbers in @strtod@ syntax, optionally followed by a unit . | |
2111 | */ | |
2112 | ||
2113 | static int parse_duration(union tvec_regval *rv, | |
2114 | const struct tvec_regdef *rd, | |
2115 | struct tvec_state *tv) | |
2116 | { | |
2117 | const struct duration_unit *u; | |
2118 | const char *q; | |
adec5584 | 2119 | dstr d = DSTR_INIT; |
814e42ff MW |
2120 | double t; |
2121 | int rc; | |
2122 | ||
adec5584 | 2123 | if (tvec_readword(tv, &d, 0, ";", "duration")) { rc = -1; goto end; } |
814e42ff MW |
2124 | if (parse_floating(&t, &q, d.buf, |
2125 | rd->arg.p ? rd->arg.p : &tvflt_nonneg, tv)) | |
2126 | { rc = -1; goto end; } | |
2127 | ||
adec5584 MW |
2128 | if (!*q) tvec_readword(tv, &d, &q, ";", 0); |
2129 | if (*q) { | |
814e42ff MW |
2130 | for (u = duration_units; u->unit; u++) |
2131 | if (STRCMP(q, ==, u->unit)) { t *= u->scale; goto found_unit; } | |
2132 | rc = tvec_syntax(tv, *q, "end-of-line"); goto end; | |
2133 | found_unit:; | |
2134 | } | |
2135 | ||
2136 | if (tvec_flushtoeol(tv, 0)) { rc = -1; goto end; } | |
2137 | rv->f = t; rc = 0; | |
2138 | end: | |
2139 | dstr_destroy(&d); | |
2140 | return (rc); | |
2141 | } | |
2142 | ||
2143 | /* --- @dump_duration@ --- * | |
2144 | * | |
2145 | * Arguments: @const union tvec_regval *rv@ = register value | |
2146 | * @const struct tvec_regdef *rd@ = register definition | |
2147 | * @unsigned style@ = output style (@TVSF_...@) | |
2148 | * @const struct gprintf_ops *gops@, @void *gp@ = format output | |
2149 | * | |
2150 | * Returns: --- | |
2151 | * | |
2152 | * Use: Dump a register value to the format output. | |
2153 | * | |
2154 | * Durations are dumped as a human-palatable scaled value with | |
2155 | * unit, and, if compact style is not requested, as a raw number | |
2156 | * of seconds at full precision as a comment. | |
2157 | */ | |
2158 | ||
2159 | static void dump_duration(const union tvec_regval *rv, | |
2160 | const struct tvec_regdef *rd, | |
2161 | unsigned style, | |
2162 | const struct gprintf_ops *gops, void *go) | |
2163 | { | |
2164 | const struct duration_unit *u; | |
2165 | double t = rv->f; | |
2166 | ||
5c0f2e08 MW |
2167 | if (style&TVSF_RAW) { |
2168 | gprintf(gops, go, "duration:"); | |
814e42ff | 2169 | format_floating(gops, go, rv->f); |
5c0f2e08 MW |
2170 | gprintf(gops, go, "s"); |
2171 | } else { | |
2172 | if (!t) u = 0; | |
2173 | else { | |
2174 | for (u = duration_units; u->scale > t && u[1].unit; u++); | |
2175 | t /= u->scale; | |
2176 | } | |
2177 | gprintf(gops, go, "%.4g %s", t, u ? u->unit : "s"); | |
2178 | ||
2179 | if (!(style&TVSF_COMPACT)) { | |
2180 | gprintf(gops, go, "; = "); | |
2181 | format_floating(gops, go, rv->f); | |
2182 | gprintf(gops, go, " s"); | |
2183 | } | |
814e42ff MW |
2184 | } |
2185 | } | |
2186 | ||
2187 | /* Duration type definition. */ | |
2188 | const struct tvec_regty tvty_duration = { | |
2189 | init_float, trivial_release, eq_float, | |
2190 | tobuf_float, frombuf_float, | |
2191 | parse_duration, dump_duration | |
2192 | }; | |
2193 | ||
c4ccbbf9 MW |
2194 | /* --- @tvec_claimeqish_duration@ --- * |
2195 | * | |
2196 | * Arguments: @struct tvec_state *tv@ = test-vector state | |
b1a20bee | 2197 | * @double t0, t1@ = two durations |
c4ccbbf9 MW |
2198 | * @unsigned f@ = flags (@TVFF_...@) |
2199 | * @double delta@ = maximum tolerable difference | |
2200 | * @const char *file@, @unsigned @lno@ = calling file and line | |
2201 | * @const char *expr@ = the expression to quote on failure | |
2202 | * | |
2203 | * Returns: Nonzero if @t0@ and @t1@ are sufficiently close, otherwise | |
2204 | * zero. | |
2205 | * | |
2206 | * Use: Check that values of @t0@ and @t1@ are sufficiently close. | |
2207 | * This is essentially the same as @tvec_claimeqish_float@, only | |
2208 | * it dumps the values as durations on a mismatch. | |
2209 | */ | |
2210 | ||
2211 | int tvec_claimeqish_duration(struct tvec_state *tv, | |
2212 | double t0, double t1, unsigned f, double delta, | |
2213 | const char *file, unsigned lno, | |
2214 | const char *expr) | |
2215 | { | |
2216 | struct tvec_floatinfo fi; | |
2217 | union tvec_misc arg; | |
2218 | ||
2219 | fi.f = f; fi.min = fi.max = 0.0; fi.delta = delta; arg.p = &fi; | |
2220 | tv->out[0].v.f = t0; tv->in[0].v.f = t1; | |
2221 | return (tvec_claimeq(tv, &tvty_duration, &arg, file, lno, expr)); | |
2222 | } | |
2223 | ||
2224 | /* --- @tvec_claimeq_duration@ --- * | |
2225 | * | |
2226 | * Arguments: @struct tvec_state *tv@ = test-vector state | |
2227 | * @double t0, t1@ = two durations | |
2228 | * @const char *file@, @unsigned @lno@ = calling file and line | |
2229 | * @const char *expr@ = the expression to quote on failure | |
2230 | * | |
2231 | * Returns: Nonzero if @t0@ and @t1@ are identical, otherwise zero. | |
2232 | * | |
2233 | * Use: Check that values of @t0@ and @t1@ are identical. The | |
2234 | * function is exactly equivalent to @tvec_claimeqish_duration@ | |
2235 | * with @f == TVFF_EXACT@. | |
2236 | */ | |
2237 | ||
2238 | int tvec_claimeq_duration(struct tvec_state *tv, | |
2239 | double t0, double t1, | |
2240 | const char *file, unsigned lno, | |
2241 | const char *expr) | |
2242 | { | |
2243 | return (tvec_claimeqish_duration(tv, t0, t1, TVFF_EXACT, 0.0, | |
2244 | file, lno, expr)); | |
2245 | } | |
2246 | ||
b64eb60f MW |
2247 | /*----- Enumerations ------------------------------------------------------*/ |
2248 | ||
c81c35df MW |
2249 | /* --- @init_tenum@ --- * |
2250 | * | |
2251 | * Arguments: @union tvec_regval *rv@ = register value | |
2252 | * @const struct tvec_regdef *rd@ = register definition | |
2253 | * | |
2254 | * Returns: --- | |
2255 | * | |
2256 | * Use: Initialize a register value. | |
2257 | * | |
2258 | * Integer and floating-point enumeration values are initialized | |
2259 | * as their underlying representations. Pointer enumerations | |
2260 | * are initialized to %|#nil|%. | |
2261 | */ | |
2262 | ||
3efcfd2d MW |
2263 | #define init_ienum init_int |
2264 | #define init_uenum init_uint | |
2265 | #define init_fenum init_float | |
c81c35df | 2266 | |
3efcfd2d MW |
2267 | static void init_penum(union tvec_regval *rv, const struct tvec_regdef *rd) |
2268 | { rv->p = 0; } | |
b64eb60f | 2269 | |
c81c35df MW |
2270 | /* --- @eq_tenum@ --- * |
2271 | * | |
2272 | * Arguments: @const union tvec_regval *rv0, *rv1@ = register values | |
2273 | * @const struct tvec_regdef *rd@ = register definition | |
2274 | * | |
2275 | * Returns: Nonzero if the values are equal, zero if unequal | |
2276 | * | |
2277 | * Use: Compare register values for equality. | |
2278 | * | |
2279 | * Integer and floating-point enumeration values are compared as | |
2280 | * their underlying representations; in particular, floating- | |
2281 | * point enumerations may compare equal if their absolute or | |
2282 | * relative difference is sufficiently small. Pointer | |
2283 | * enumerations are compared as pointers. | |
2284 | */ | |
2285 | ||
3efcfd2d MW |
2286 | #define eq_ienum eq_int |
2287 | #define eq_uenum eq_uint | |
c81c35df | 2288 | |
3efcfd2d MW |
2289 | static int eq_fenum(const union tvec_regval *rv0, |
2290 | const union tvec_regval *rv1, | |
2291 | const struct tvec_regdef *rd) | |
b64eb60f | 2292 | { |
3efcfd2d MW |
2293 | const struct tvec_fenuminfo *ei = rd->arg.p; |
2294 | return (eqish_floating_p(rv0->f, rv1->f, ei->fi)); | |
b64eb60f | 2295 | } |
c81c35df | 2296 | |
3efcfd2d MW |
2297 | static int eq_penum(const union tvec_regval *rv0, |
2298 | const union tvec_regval *rv1, | |
2299 | const struct tvec_regdef *rd) | |
2300 | { return (rv0->p == rv1->p); } | |
b64eb60f | 2301 | |
c81c35df MW |
2302 | /* --- @tobuf_tenum@ --- * |
2303 | * | |
2304 | * Arguments: @buf *b@ = buffer | |
2305 | * @const union tvec_regval *rv@ = register value | |
2306 | * @const struct tvec_regdef *rd@ = register definition | |
2307 | * | |
2308 | * Returns: Zero on success, %$-1$% on failure. | |
2309 | * | |
2310 | * Use: Serialize a register value to a buffer. | |
2311 | * | |
2312 | * Integer and floating-point enumeration values are serialized | |
2313 | * as their underlying representations. Pointer enumerations | |
2314 | * are serialized as the signed integer index into the | |
2315 | * association table; %|#nil|% serializes as %$-1$%, and | |
2316 | * unrecognized pointers cause failure. | |
2317 | */ | |
2318 | ||
3efcfd2d MW |
2319 | #define tobuf_ienum tobuf_int |
2320 | #define tobuf_uenum tobuf_uint | |
2321 | #define tobuf_fenum tobuf_float | |
c81c35df | 2322 | |
3efcfd2d MW |
2323 | static int tobuf_penum(buf *b, const union tvec_regval *rv, |
2324 | const struct tvec_regdef *rd) | |
b64eb60f | 2325 | { |
3efcfd2d | 2326 | const struct tvec_penuminfo *pei = rd->arg.p; |
e63124bc MW |
2327 | const struct tvec_passoc *pa; |
2328 | long i; | |
b64eb60f | 2329 | |
3efcfd2d MW |
2330 | for (pa = pei->av, i = 0; pa->tag; pa++, i++) |
2331 | if (pa->p == rv->p) goto found; | |
2332 | if (!rv->p) i = -1; | |
2333 | else return (-1); | |
2334 | found: | |
2335 | return (signed_to_buf(b, i)); | |
b64eb60f MW |
2336 | } |
2337 | ||
c81c35df MW |
2338 | /* --- @frombuf_tenum@ --- * |
2339 | * | |
2340 | * Arguments: @buf *b@ = buffer | |
2341 | * @union tvec_regval *rv@ = register value | |
2342 | * @const struct tvec_regdef *rd@ = register definition | |
2343 | * | |
2344 | * Returns: Zero on success, %$-1$% on failure. | |
2345 | * | |
2346 | * Use: Deserialize a register value from a buffer. | |
2347 | * | |
2348 | * Integer and floating-point enumeration values are serialized | |
2349 | * as their underlying representations. Pointer enumerations | |
2350 | * are serialized as the signed integer index into the | |
2351 | * association table; %|#nil|% serializes as %$-1$%; out-of- | |
2352 | * range indices cause failure. | |
2353 | */ | |
2354 | ||
3efcfd2d MW |
2355 | #define frombuf_ienum frombuf_int |
2356 | #define frombuf_uenum frombuf_uint | |
2357 | #define frombuf_fenum frombuf_float | |
2358 | static int frombuf_penum(buf *b, union tvec_regval *rv, | |
b64eb60f MW |
2359 | const struct tvec_regdef *rd) |
2360 | { | |
3efcfd2d | 2361 | const struct tvec_penuminfo *pei = rd->arg.p; |
e63124bc MW |
2362 | const struct tvec_passoc *pa; |
2363 | long i, n; | |
b64eb60f | 2364 | |
3efcfd2d MW |
2365 | for (pa = pei->av, n = 0; pa->tag; pa++, n++); |
2366 | if (signed_from_buf(b, &i)) return (-1); | |
b1a20bee | 2367 | if (0 <= i && i < n) rv->p = UNCONST(void, pei->av[i].p); |
3efcfd2d | 2368 | else if (i == -1) rv->p = 0; |
adec5584 | 2369 | else { buf_break(b); return (-1); } |
3efcfd2d | 2370 | return (0); |
b64eb60f MW |
2371 | } |
2372 | ||
c81c35df MW |
2373 | /* --- @parse_tenum@ --- * |
2374 | * | |
2375 | * Arguments: @union tvec_regval *rv@ = register value | |
2376 | * @const struct tvec_regdef *rd@ = register definition | |
2377 | * @struct tvec_state *tv@ = test-vector state | |
2378 | * | |
2379 | * Returns: Zero on success, %$-1$% on error. | |
2380 | * | |
2381 | * Use: Parse a register value from an input file. | |
2382 | * | |
2383 | * An enumerated value may be given by name or as a literal | |
2384 | * value. For enumerations based on numeric types, the literal | |
2385 | * values can be written in the same syntax as the underlying | |
2386 | * values. For enumerations based on pointers, the only | |
2387 | * permitted literal is %|#nil|%, which denotes a null pointer. | |
2388 | */ | |
2389 | ||
3efcfd2d MW |
2390 | #define DEFPARSE_ENUM(tag_, ty, slot) \ |
2391 | static int parse_##slot##enum(union tvec_regval *rv, \ | |
2392 | const struct tvec_regdef *rd, \ | |
2393 | struct tvec_state *tv) \ | |
2394 | { \ | |
2395 | const struct tvec_##slot##enuminfo *ei = rd->arg.p; \ | |
2396 | const struct tvec_##slot##assoc *a; \ | |
2397 | dstr d = DSTR_INIT; \ | |
2398 | int rc; \ | |
2399 | \ | |
5c0f2e08 MW |
2400 | if (tvec_readword(tv, &d, 0, \ |
2401 | ";", "%s tag or " LITSTR_##tag_, ei->name)) \ | |
3efcfd2d MW |
2402 | { rc = -1; goto end; } \ |
2403 | for (a = ei->av; a->tag; a++) \ | |
2404 | if (STRCMP(a->tag, ==, d.buf)) { FOUND_##tag_ goto done; } \ | |
2405 | MISSING_##tag_ \ | |
2406 | done: \ | |
2407 | if (tvec_flushtoeol(tv, 0)) { rc = -1; goto end; } \ | |
2408 | rc = 0; \ | |
2409 | end: \ | |
2410 | dstr_destroy(&d); \ | |
2411 | return (rc); \ | |
2412 | } | |
b64eb60f | 2413 | |
3efcfd2d MW |
2414 | #define LITSTR_INT "literal signed integer" |
2415 | #define FOUND_INT rv->i = a->i; | |
2416 | #define MISSING_INT if (parse_signed(&rv->i, d.buf, ei->ir, tv)) \ | |
2417 | { rc = -1; goto end; } | |
2418 | ||
2419 | #define LITSTR_UINT "literal unsigned integer" | |
2420 | #define FOUND_UINT rv->u = a->u; | |
2421 | #define MISSING_UINT if (parse_unsigned(&rv->u, d.buf, ei->ur, tv)) \ | |
2422 | { rc = -1; goto end; } | |
2423 | ||
2424 | #define LITSTR_FLT "literal floating-point number, " \ | |
2425 | "`#-inf', `#+inf', or `#nan'" | |
2426 | #define FOUND_FLT rv->f = a->f; | |
814e42ff | 2427 | #define MISSING_FLT if (parse_floating(&rv->f, 0, d.buf, ei->fi, tv)) \ |
3efcfd2d MW |
2428 | { rc = -1; goto end; } |
2429 | ||
2430 | #define LITSTR_PTR "`#nil'" | |
b1a20bee | 2431 | #define FOUND_PTR rv->p = UNCONST(void, a->p); |
3efcfd2d MW |
2432 | #define MISSING_PTR if (STRCMP(d.buf, ==, "#nil")) \ |
2433 | rv->p = 0; \ | |
2434 | else { \ | |
2435 | tvec_error(tv, "unknown `%s' value `%s'", \ | |
2436 | ei->name, d.buf); \ | |
2437 | rc = -1; goto end; \ | |
2438 | } | |
2439 | ||
2440 | TVEC_MISCSLOTS(DEFPARSE_ENUM) | |
2441 | ||
2442 | #undef LITSTR_INT | |
2443 | #undef FOUND_INT | |
2444 | #undef MISSING_INT | |
2445 | ||
2446 | #undef LITSTR_UINT | |
2447 | #undef FOUND_UINT | |
2448 | #undef MISSING_UINT | |
2449 | ||
2450 | #undef LITSTR_FLT | |
2451 | #undef FOUND_FLT | |
2452 | #undef MISSING_FLT | |
2453 | ||
2454 | #undef LITSTR_PTR | |
2455 | #undef FOUND_PTR | |
2456 | #undef MISSING_PTR | |
2457 | ||
2458 | #undef DEFPARSE_ENUM | |
2459 | ||
c81c35df MW |
2460 | /* --- @dump_tenum@ --- * |
2461 | * | |
2462 | * Arguments: @const union tvec_regval *rv@ = register value | |
2463 | * @const struct tvec_regdef *rd@ = register definition | |
2464 | * @unsigned style@ = output style (@TVSF_...@) | |
2465 | * @const struct gprintf_ops *gops@, @void *gp@ = format output | |
2466 | * | |
2467 | * Returns: --- | |
2468 | * | |
2469 | * Use: Dump a register value to the format output. | |
2470 | * | |
2471 | * Enumeration values are dumped as their symbolic names, if | |
2472 | * possible, with the underlying values provided as a comment | |
2473 | * unless compact output is requested, as for the underlying | |
2474 | * representation. A null pointer is printed as %|#nil|%; | |
2475 | * non-null pointers are printed as %|#<TYPE PTR>|%, with the | |
2476 | * enumeration TYPE and the raw pointer PTR printed with the | |
2477 | * system's %|%p|% format specifier. | |
2478 | */ | |
2479 | ||
2480 | ||
3efcfd2d MW |
2481 | #define DEFDUMP_ENUM(tag_, ty, slot) \ |
2482 | static void dump_##slot##enum(const union tvec_regval *rv, \ | |
2483 | const struct tvec_regdef *rd, \ | |
2484 | unsigned style, \ | |
2485 | const struct gprintf_ops *gops, void *go) \ | |
2486 | { \ | |
2487 | const struct tvec_##slot##enuminfo *ei = rd->arg.p; \ | |
2488 | const struct tvec_##slot##assoc *a; \ | |
2489 | \ | |
5c0f2e08 | 2490 | if (style&TVSF_RAW) gprintf(gops, go, #slot "enum/%s:", ei->name); \ |
3efcfd2d MW |
2491 | for (a = ei->av; a->tag; a++) \ |
2492 | if (rv->slot == a->slot) { \ | |
2493 | gprintf(gops, go, "%s", a->tag); \ | |
2494 | if (style&TVSF_COMPACT) return; \ | |
2495 | gprintf(gops, go, " ; = "); break; \ | |
2496 | } \ | |
2497 | \ | |
2498 | PRINTRAW_##tag_ \ | |
b64eb60f MW |
2499 | } |
2500 | ||
3efcfd2d | 2501 | #define MAYBE_PRINT_EXTRA \ |
c81c35df | 2502 | if (style&TVSF_COMPACT) /* nothing to do */; \ |
3efcfd2d MW |
2503 | else if (!a->tag) { gprintf(gops, go, " ; = "); goto _extra; } \ |
2504 | else if (1) { gprintf(gops, go, " = "); goto _extra; } \ | |
2505 | else _extra: | |
b64eb60f | 2506 | |
3efcfd2d MW |
2507 | #define PRINTRAW_INT gprintf(gops, go, "%ld", rv->i); \ |
2508 | MAYBE_PRINT_EXTRA { \ | |
2509 | format_signed_hex(gops, go, rv->i); \ | |
2510 | maybe_format_signed_char(gops, go, rv->i); \ | |
2511 | } | |
b64eb60f | 2512 | |
3efcfd2d MW |
2513 | #define PRINTRAW_UINT gprintf(gops, go, "%lu", rv->u); \ |
2514 | MAYBE_PRINT_EXTRA { \ | |
2515 | format_unsigned_hex(gops, go, rv->u); \ | |
2516 | maybe_format_unsigned_char(gops, go, rv->u); \ | |
2517 | } | |
2518 | ||
2519 | #define PRINTRAW_FLT format_floating(gops, go, rv->f); | |
2520 | ||
2521 | #define PRINTRAW_PTR if (!rv->p) gprintf(gops, go, "#nil"); \ | |
e63124bc | 2522 | else gprintf(gops, go, "#<%s %p>", ei->name, rv->p); |
b64eb60f | 2523 | |
3efcfd2d | 2524 | TVEC_MISCSLOTS(DEFDUMP_ENUM) |
b64eb60f | 2525 | |
3efcfd2d MW |
2526 | #undef PRINTRAW_INT |
2527 | #undef PRINTRAW_UINT | |
2528 | #undef PRINTRAW_FLT | |
2529 | #undef PRINTRAW_PTR | |
2530 | ||
2531 | #undef MAYBE_PRINT_EXTRA | |
2532 | #undef DEFDUMP_ENUM | |
2533 | ||
c81c35df | 2534 | /* Enumeration type definitions. */ |
3efcfd2d MW |
2535 | #define DEFTY_ENUM(tag, ty, slot) \ |
2536 | const struct tvec_regty tvty_##slot##enum = { \ | |
2537 | init_##slot##enum, trivial_release, eq_##slot##enum, \ | |
2538 | tobuf_##slot##enum, frombuf_##slot##enum, \ | |
2539 | parse_##slot##enum, dump_##slot##enum \ | |
2540 | }; | |
2541 | TVEC_MISCSLOTS(DEFTY_ENUM) | |
2542 | #undef DEFTY_ENUM | |
b64eb60f | 2543 | |
c81c35df | 2544 | /* Predefined enumeration types. */ |
e63124bc MW |
2545 | static const struct tvec_iassoc bool_assoc[] = { |
2546 | { "nil", 0 }, | |
2547 | { "false", 0 }, | |
2548 | { "f", 0 }, | |
2549 | { "no", 0 }, | |
2550 | { "n", 0 }, | |
2551 | { "off", 0 }, | |
2552 | ||
2553 | { "t", 1 }, | |
2554 | { "true", 1 }, | |
2555 | { "yes", 1 }, | |
2556 | { "y", 1 }, | |
2557 | { "on", 1 }, | |
2558 | ||
20ba6b0b | 2559 | TVEC_ENDENUM |
e63124bc MW |
2560 | }; |
2561 | ||
2562 | const struct tvec_ienuminfo tvenum_bool = | |
3efcfd2d | 2563 | { "bool", bool_assoc, &tvrange_int }; |
e63124bc | 2564 | |
20ba6b0b MW |
2565 | static const struct tvec_iassoc cmp_assoc[] = { |
2566 | { "<", -1 }, | |
2567 | { "less", -1 }, | |
2568 | { "lt", -1 }, | |
2569 | ||
2570 | { "=", 0 }, | |
2571 | { "equal", 0 }, | |
2572 | { "eq", 0 }, | |
2573 | ||
2574 | { ">", +1 }, | |
2575 | { "greater", +1 }, | |
2576 | { "gt", +1 }, | |
2577 | ||
2578 | TVEC_ENDENUM | |
2579 | }; | |
2580 | ||
2581 | const struct tvec_ienuminfo tvenum_cmp = | |
2582 | { "cmp", cmp_assoc, &tvrange_int }; | |
2583 | ||
67b5031e MW |
2584 | /* --- @tvec_claimeq_tenum@ --- * |
2585 | * | |
2586 | * Arguments: @struct tvec_state *tv@ = test-vector state | |
2587 | * @const struct tvec_typeenuminfo *ei@ = enumeration type info | |
2588 | * @ty t0, t1@ = two values | |
2589 | * @const char *file@, @unsigned @lno@ = calling file and line | |
2590 | * @const char *expr@ = the expression to quote on failure | |
2591 | * | |
2592 | * Returns: Nonzero if @t0@ and @t1@ are equal, otherwise zero. | |
2593 | * | |
2594 | * Use: Check that values of @t0@ and @t1@ are equal. As for | |
2595 | * @tvec_claim@ above, a test case is automatically begun and | |
2596 | * ended if none is already underway. If the values are | |
2597 | * unequal, then @tvec_fail@ is called, quoting @expr@, and the | |
2598 | * mismatched values are dumped: @t0@ is printed as the output | |
2599 | * value and @t1@ is printed as the input reference. | |
2600 | */ | |
2601 | ||
b64eb60f | 2602 | #define DEFCLAIM(tag, ty, slot) \ |
e63124bc MW |
2603 | int tvec_claimeq_##slot##enum \ |
2604 | (struct tvec_state *tv, \ | |
2605 | const struct tvec_##slot##enuminfo *ei, ty e0, ty e1, \ | |
2606 | const char *file, unsigned lno, const char *expr) \ | |
b64eb60f MW |
2607 | { \ |
2608 | union tvec_misc arg; \ | |
2609 | \ | |
b64eb60f | 2610 | arg.p = ei; \ |
3efcfd2d MW |
2611 | tv->out[0].v.slot = GET_##tag(e0); \ |
2612 | tv->in[0].v.slot = GET_##tag(e1); \ | |
2613 | return (tvec_claimeq(tv, &tvty_##slot##enum, &arg, \ | |
2614 | file, lno, expr)); \ | |
b64eb60f MW |
2615 | } |
2616 | #define GET_INT(e) (e) | |
2617 | #define GET_UINT(e) (e) | |
e63124bc | 2618 | #define GET_FLT(e) (e) |
b1a20bee | 2619 | #define GET_PTR(e) (UNCONST(void, (e))) |
b64eb60f MW |
2620 | TVEC_MISCSLOTS(DEFCLAIM) |
2621 | #undef DEFCLAIM | |
2622 | #undef GET_INT | |
2623 | #undef GET_UINT | |
e63124bc | 2624 | #undef GET_FLT |
b64eb60f MW |
2625 | #undef GET_PTR |
2626 | ||
2627 | /*----- Flag types --------------------------------------------------------*/ | |
2628 | ||
c81c35df MW |
2629 | /* Flag types are initialized, compared, and serialized as unsigned |
2630 | * integers. | |
2631 | */ | |
2632 | ||
2633 | /* --- @parse_flags@ --- * | |
2634 | * | |
2635 | * Arguments: @union tvec_regval *rv@ = register value | |
2636 | * @const struct tvec_regdef *rd@ = register definition | |
2637 | * @struct tvec_state *tv@ = test-vector state | |
2638 | * | |
2639 | * Returns: Zero on success, %$-1$% on error. | |
2640 | * | |
2641 | * Use: Parse a register value from an input file. | |
2642 | * | |
2643 | * The input syntax is a sequence of items separated by `|' | |
2644 | * signs. Each item may be the symbolic name of a field value, | |
2645 | * or a literal unsigned integer. The masks associated with the | |
2646 | * given symbolic names must be disjoint. The resulting | |
2647 | * numerical value is simply the bitwise OR of the given values. | |
2648 | */ | |
2649 | ||
882a39c1 MW |
2650 | static int parse_flags(union tvec_regval *rv, const struct tvec_regdef *rd, |
2651 | struct tvec_state *tv) | |
b64eb60f MW |
2652 | { |
2653 | const struct tvec_flaginfo *fi = rd->arg.p; | |
2654 | const struct tvec_flag *f; | |
2655 | unsigned long m = 0, v = 0, t; | |
2656 | dstr d = DSTR_INIT; | |
882a39c1 | 2657 | int ch, rc; |
b64eb60f MW |
2658 | |
2659 | for (;;) { | |
c81c35df MW |
2660 | |
2661 | /* Read the next item. */ | |
882a39c1 | 2662 | DRESET(&d); |
5c0f2e08 | 2663 | if (tvec_readword(tv, &d, 0, "|;", "%s flag name or integer", fi->name)) |
882a39c1 | 2664 | { rc = -1; goto end; } |
b64eb60f | 2665 | |
c81c35df | 2666 | /* Try to find a matching entry in the table. */ |
b64eb60f MW |
2667 | for (f = fi->fv; f->tag; f++) |
2668 | if (STRCMP(f->tag, ==, d.buf)) { | |
882a39c1 MW |
2669 | if (m&f->m) |
2670 | { tvec_error(tv, "colliding flag setting"); rc = -1; goto end; } | |
2671 | else | |
2672 | { m |= f->m; v |= f->v; goto next; } | |
b64eb60f MW |
2673 | } |
2674 | ||
c81c35df | 2675 | /* Otherwise, try to parse it as a raw integer. */ |
e63124bc MW |
2676 | if (parse_unsigned(&t, d.buf, fi->range, tv)) |
2677 | { rc = -1; goto end; } | |
882a39c1 | 2678 | v |= t; |
c81c35df | 2679 | |
b64eb60f | 2680 | next: |
c81c35df MW |
2681 | /* Advance to the next token. If it's a separator then consume it, and |
2682 | * go round again. Otherwise we stop here. | |
2683 | */ | |
b64eb60f | 2684 | if (tvec_nexttoken(tv)) break; |
882a39c1 MW |
2685 | ch = getc(tv->fp); |
2686 | if (ch != '|') { tvec_syntax(tv, ch, "`|'"); rc = -1; goto end; } | |
5c0f2e08 MW |
2687 | if (tvec_nexttoken(tv)) { |
2688 | tvec_syntax(tv, '\n', "%s flag name or integer", fi->name); | |
2689 | rc = -1; goto end; | |
2690 | } | |
b64eb60f | 2691 | } |
c81c35df MW |
2692 | |
2693 | /* Done. */ | |
2694 | rv->u = v; rc = 0; | |
882a39c1 MW |
2695 | end: |
2696 | dstr_destroy(&d); | |
2697 | return (rc); | |
b64eb60f MW |
2698 | } |
2699 | ||
c81c35df MW |
2700 | /* --- @dump_flags@ --- * |
2701 | * | |
2702 | * Arguments: @const union tvec_regval *rv@ = register value | |
2703 | * @const struct tvec_regdef *rd@ = register definition | |
2704 | * @unsigned style@ = output style (@TVSF_...@) | |
2705 | * @const struct gprintf_ops *gops@, @void *gp@ = format output | |
2706 | * | |
2707 | * Returns: --- | |
2708 | * | |
2709 | * Use: Dump a register value to the format output. | |
2710 | * | |
2711 | * The table of symbolic names and their associated values and | |
2712 | * masks is repeatedly scanned, in order, to find disjoint | |
2713 | * matches -- i.e., entries whose value matches the target value | |
2714 | * in the bit positions indicated by the mask, and whose mask | |
2715 | * doesn't overlap with any previously found matches; the names | |
2716 | * are then output, separated by `|'. Any remaining nonzero | |
2717 | * bits not covered by any of the matching masks are output as a | |
2718 | * single literal integer, in hex. | |
2719 | * | |
2720 | * Unless compact output is requested, or no symbolic names were | |
2721 | * found, the raw numeric value is also printed in hex, as a | |
2722 | * comment. | |
2723 | */ | |
2724 | ||
b64eb60f MW |
2725 | static void dump_flags(const union tvec_regval *rv, |
2726 | const struct tvec_regdef *rd, | |
e63124bc MW |
2727 | unsigned style, |
2728 | const struct gprintf_ops *gops, void *go) | |
b64eb60f MW |
2729 | { |
2730 | const struct tvec_flaginfo *fi = rd->arg.p; | |
2731 | const struct tvec_flag *f; | |
c81c35df | 2732 | unsigned long m = ~0ul, v = rv->u; |
b64eb60f MW |
2733 | const char *sep; |
2734 | ||
5c0f2e08 MW |
2735 | if (style&TVSF_RAW) gprintf(gops, go, "flags/%s:", fi->name); |
2736 | ||
b64eb60f MW |
2737 | for (f = fi->fv, sep = ""; f->tag; f++) |
2738 | if ((m&f->m) && (v&f->m) == f->v) { | |
e63124bc | 2739 | gprintf(gops, go, "%s%s", sep, f->tag); m &= ~f->m; |
b64eb60f MW |
2740 | sep = style&TVSF_COMPACT ? "|" : " | "; |
2741 | } | |
2742 | ||
e63124bc | 2743 | if (v&m) gprintf(gops, go, "%s0x%0*lx", sep, hex_width(v), v&m); |
b1a20bee | 2744 | else if (!v && m == ~0ul) gprintf(gops, go, "0"); |
b64eb60f | 2745 | |
b1a20bee | 2746 | if (!(style&(TVSF_COMPACT | TVSF_RAW))) |
e63124bc | 2747 | gprintf(gops, go, " ; = 0x%0*lx", hex_width(rv->u), rv->u); |
b64eb60f MW |
2748 | } |
2749 | ||
c81c35df | 2750 | /* Flags type definition. */ |
b64eb60f | 2751 | const struct tvec_regty tvty_flags = { |
3efcfd2d | 2752 | init_uint, trivial_release, eq_uint, |
b64eb60f MW |
2753 | tobuf_uint, frombuf_uint, |
2754 | parse_flags, dump_flags | |
2755 | }; | |
2756 | ||
67b5031e MW |
2757 | /* --- @tvec_claimeq_flags@ --- * |
2758 | * | |
2759 | * Arguments: @struct tvec_state *tv@ = test-vector state | |
2760 | * @const struct tvec_flaginfo *fi@ = flags type info | |
2761 | * @unsigned long f0, f1@ = two values | |
2762 | * @const char *file@, @unsigned @lno@ = calling file and line | |
2763 | * @const char *expr@ = the expression to quote on failure | |
2764 | * | |
2765 | * Returns: Nonzero if @f0@ and @f1@ are equal, otherwise zero. | |
2766 | * | |
2767 | * Use: Check that values of @f0@ and @f1@ are equal. As for | |
2768 | * @tvec_claim@ above, a test case is automatically begun and | |
2769 | * ended if none is already underway. If the values are | |
2770 | * unequal, then @tvec_fail@ is called, quoting @expr@, and the | |
2771 | * mismatched values are dumped: @f0@ is printed as the output | |
2772 | * value and @f1@ is printed as the input reference. | |
2773 | */ | |
2774 | ||
b64eb60f MW |
2775 | int tvec_claimeq_flags(struct tvec_state *tv, |
2776 | const struct tvec_flaginfo *fi, | |
2777 | unsigned long f0, unsigned long f1, | |
2778 | const char *file, unsigned lno, const char *expr) | |
2779 | { | |
2780 | union tvec_misc arg; | |
2781 | ||
3efcfd2d | 2782 | arg.p = fi; tv->out[0].v.u = f0; tv->in[0].v.u = f1; |
b64eb60f MW |
2783 | return (tvec_claimeq(tv, &tvty_flags, &arg, file, lno, expr)); |
2784 | } | |
2785 | ||
e63124bc MW |
2786 | /*----- Characters --------------------------------------------------------*/ |
2787 | ||
c81c35df MW |
2788 | /* Character values are initialized and compared as signed integers. */ |
2789 | ||
2790 | /* --- @tobuf_char@ --- * | |
2791 | * | |
2792 | * Arguments: @buf *b@ = buffer | |
2793 | * @const union tvec_regval *rv@ = register value | |
2794 | * @const struct tvec_regdef *rd@ = register definition | |
2795 | * | |
2796 | * Returns: Zero on success, %$-1$% on failure. | |
2797 | * | |
2798 | * Use: Serialize a register value to a buffer. | |
2799 | * | |
2800 | * Character values are serialized as little-endian 32-bit | |
2801 | * unsigned integers, with %|EOF|% serialized as all-bits-set. | |
2802 | */ | |
2803 | ||
e63124bc | 2804 | static int tobuf_char(buf *b, const union tvec_regval *rv, |
67b5031e | 2805 | const struct tvec_regdef *rd) |
e63124bc MW |
2806 | { |
2807 | uint32 u; | |
c81c35df | 2808 | |
e63124bc MW |
2809 | if (0 <= rv->i && rv->i <= UCHAR_MAX) u = rv->i; |
2810 | else if (rv->i == EOF) u = MASK32; | |
adec5584 | 2811 | else { buf_break(b); return (-1); } |
e63124bc MW |
2812 | return (buf_putu32l(b, u)); |
2813 | } | |
2814 | ||
c81c35df MW |
2815 | /* --- @frombuf_char@ --- * |
2816 | * | |
2817 | * Arguments: @buf *b@ = buffer | |
2818 | * @union tvec_regval *rv@ = register value | |
2819 | * @const struct tvec_regdef *rd@ = register definition | |
2820 | * | |
2821 | * Returns: Zero on success, %$-1$% on failure. | |
2822 | * | |
2823 | * Use: Deserialize a register value from a buffer. | |
2824 | * | |
2825 | * Character values are serialized as little-endian 32-bit | |
2826 | * unsigned integers, with %|EOF|% serialized as all-bits-set. | |
2827 | */ | |
2828 | ||
e63124bc | 2829 | static int frombuf_char(buf *b, union tvec_regval *rv, |
67b5031e | 2830 | const struct tvec_regdef *rd) |
e63124bc MW |
2831 | { |
2832 | uint32 u; | |
2833 | ||
2834 | if (buf_getu32l(b, &u)) return (-1); | |
2835 | if (0 <= u && u <= UCHAR_MAX) rv->i = u; | |
2836 | else if (u == MASK32) rv->i = EOF; | |
adec5584 | 2837 | else { buf_break(b); return (-1); } |
e63124bc MW |
2838 | return (0); |
2839 | } | |
2840 | ||
c81c35df MW |
2841 | /* --- @parse_char@ --- * |
2842 | * | |
2843 | * Arguments: @union tvec_regval *rv@ = register value | |
2844 | * @const struct tvec_regdef *rd@ = register definition | |
2845 | * @struct tvec_state *tv@ = test-vector state | |
2846 | * | |
2847 | * Returns: Zero on success, %$-1$% on error. | |
2848 | * | |
2849 | * Use: Parse a register value from an input file. | |
2850 | * | |
2851 | * A character value can be given by symbolic name, with a | |
2852 | * leading `%|#|%'; or a character or `%|\|%'-escape sequence, | |
2853 | * optionally in single quotes. | |
2854 | * | |
2855 | * The following escape sequences and character names are | |
2856 | * recognized. | |
2857 | * | |
2858 | * * `%|#eof|%' is the special end-of-file marker. | |
2859 | * | |
2860 | * * `%|#nul|%' is the NUL character, sometimes used to | |
2861 | * terminate strings. | |
2862 | * | |
2863 | * * `%|bell|%', `%|bel|%', `%|ding|%', or `%|\a|%' is the BEL | |
2864 | * character used to ring the terminal bell (or do some other | |
2865 | * thing to attract the user's attention). | |
2866 | * | |
2867 | * * %|#backspace|%, %|#bs|%, or %|\b|% is the backspace | |
2868 | * character, used to move the cursor backwords by one cell. | |
2869 | * | |
2870 | * * %|#escape|% %|#esc|%, or%|\e|% is the escape character, | |
2871 | * used to introduce special terminal commands. | |
2872 | * | |
2873 | * * %|#formfeed|%, %|#ff|%, or %|\f|% is the formfeed | |
2874 | * character, used to separate pages of text. | |
2875 | * | |
2876 | * * %|#newline|%, %|#linefeed|%, %|#lf|%, %|#nl|%, or %|\n|% is | |
2877 | * the newline character, used to terminate lines of text or | |
2878 | * advance the cursor to the next line (perhaps without | |
2879 | * returning it to the start of the line). | |
2880 | * | |
2881 | * * %|#return|%, %|#carriage-return|%, %|#cr|%, or %|\r|% is | |
2882 | * the carriage-return character, used to return the cursor to | |
2883 | * the start of the line. | |
2884 | * | |
2885 | * * %|#tab|%, %|#horizontal-tab|%, %|#ht|%, or %|\t|% is the | |
2886 | * tab character, used to advance the cursor to the next tab | |
2887 | * stop on the current line. | |
2888 | * | |
2889 | * * %|#vertical-tab|%, %|#vt|%, %|\v|% is the vertical tab | |
2890 | * character. | |
2891 | * | |
2892 | * * %|#space|%, %|#spc|% is the space character. | |
2893 | * | |
2894 | * * %|#delete|%, %|#del|% is the delete character, used to | |
2895 | * erase the most recent character. | |
2896 | * | |
2897 | * * %|\'|% is the single-quote character. | |
2898 | * | |
2899 | * * %|\\|% is the backslash character. | |
2900 | * | |
2901 | * * %|\"|% is the double-quote character. | |
2902 | * | |
2903 | * * %|\NNN|% or %|\{NNN}|% is the character with code NNN in | |
2904 | * octal. The NNN may be up to three digits long. | |
2905 | * | |
2906 | * * %|\xNN|% or %|\x{NN}|% is the character with code NNN in | |
2907 | * hexadecimal. | |
2908 | */ | |
2909 | ||
e63124bc MW |
2910 | static int parse_char(union tvec_regval *rv, const struct tvec_regdef *rd, |
2911 | struct tvec_state *tv) | |
2912 | { | |
2913 | dstr d = DSTR_INIT; | |
2914 | int ch, rc; | |
2915 | unsigned f = 0; | |
2916 | #define f_quote 1u | |
2917 | ||
c81c35df | 2918 | /* Inspect the character to see what we're up against. */ |
e63124bc | 2919 | ch = getc(tv->fp); |
c81c35df | 2920 | |
e63124bc | 2921 | if (ch == '#') { |
c81c35df MW |
2922 | /* It looks like a special token. Push the `%|#|%' back and fetch the |
2923 | * whole word. If there's just the `%|#|%' after all, then treat it as | |
2924 | * literal. | |
2925 | */ | |
2926 | ||
e63124bc | 2927 | ungetc(ch, tv->fp); |
adec5584 MW |
2928 | if (tvec_readword(tv, &d, 0, ";", "character name")) |
2929 | { rc = -1; goto end; } | |
c81c35df MW |
2930 | if (STRCMP(d.buf, !=, "#")) { |
2931 | if (read_charname(&ch, d.buf, RCF_EOFOK)) { | |
2932 | rc = tvec_error(tv, "unknown character name `%s'", d.buf); | |
2933 | goto end; | |
2934 | } | |
2935 | if (tvec_flushtoeol(tv, 0)) { rc = -1; goto end; } | |
2936 | rv->i = ch; rc = 0; goto end; | |
e63124bc | 2937 | } |
e63124bc MW |
2938 | } |
2939 | ||
c81c35df MW |
2940 | /* If this is a single quote then we expect to see a matching one later, |
2941 | * and we should process backslash escapes. Get the next character and see | |
2942 | * what happens. | |
2943 | */ | |
e63124bc | 2944 | if (ch == '\'') { f |= f_quote; ch = getc(tv->fp); } |
c81c35df MW |
2945 | |
2946 | /* Main character dispatch. */ | |
e63124bc | 2947 | switch (ch) { |
c81c35df | 2948 | |
67b5031e | 2949 | case ';': |
c81c35df | 2950 | /* Unquoted, semicolon begins a comment. */ |
67b5031e | 2951 | if (!(f&f_quote)) { rc = tvec_syntax(tv, ch, "character"); goto end; } |
c81c35df MW |
2952 | else goto plain; |
2953 | ||
67b5031e | 2954 | case '\n': |
c81c35df MW |
2955 | /* A newline. If we saw a single quote, then treat that as literal. |
2956 | * Otherwise this is an error. | |
2957 | */ | |
2958 | if (!(f&f_quote)) goto nochar; | |
2959 | else { f &= ~f_quote; ungetc(ch, tv->fp); ch = '\''; goto plain; } | |
2960 | ||
67b5031e | 2961 | case EOF: |
c81c35df MW |
2962 | /* End-of-file. Similar to newline, but with slightly different |
2963 | * effects on the parse state. | |
2964 | */ | |
2965 | if (!(f&f_quote)) goto nochar; | |
2966 | else { f &= ~f_quote; ch = '\''; goto plain; } | |
2967 | ||
2968 | case '\'': nochar: | |
2969 | /* A single quote. This must be the second of a pair, and there should | |
2970 | * have been a character or escape sequence between them. | |
2971 | */ | |
e63124bc | 2972 | rc = tvec_syntax(tv, ch, "character"); goto end; |
c81c35df | 2973 | |
e63124bc | 2974 | case '\\': |
c81c35df | 2975 | /* A backslash. Read a character escape. */ |
67b5031e | 2976 | if (read_charesc(&ch, tv)) return (-1); |
c81c35df | 2977 | |
e63124bc | 2978 | default: plain: |
c81c35df | 2979 | /* Anything else. Treat as literal. */ |
e63124bc MW |
2980 | rv->i = ch; break; |
2981 | } | |
c81c35df MW |
2982 | |
2983 | /* If we saw an opening quote, then expect the closing quote. */ | |
e63124bc MW |
2984 | if (f&f_quote) { |
2985 | ch = getc(tv->fp); | |
2986 | if (ch != '\'') { rc = tvec_syntax(tv, ch, "`''"); goto end; } | |
2987 | } | |
c81c35df MW |
2988 | |
2989 | /* Done. */ | |
e63124bc MW |
2990 | if (tvec_flushtoeol(tv, 0)) { rc = -1; goto end; } |
2991 | rc = 0; | |
2992 | end: | |
2993 | dstr_destroy(&d); | |
2994 | return (rc); | |
2995 | ||
2996 | #undef f_quote | |
2997 | } | |
2998 | ||
c81c35df MW |
2999 | /* --- @dump_char@ --- * |
3000 | * | |
3001 | * Arguments: @const union tvec_regval *rv@ = register value | |
3002 | * @const struct tvec_regdef *rd@ = register definition | |
3003 | * @unsigned style@ = output style (@TVSF_...@) | |
3004 | * @const struct gprintf_ops *gops@, @void *gp@ = format output | |
3005 | * | |
3006 | * Returns: --- | |
3007 | * | |
3008 | * Use: Dump a register value to the format output. | |
3009 | * | |
3010 | * Character values are dumped as their symbolic names, if any, | |
3011 | * or as a character or escape sequence within single quotes | |
3012 | * (which may be omitted in compact style). If compact output | |
3013 | * is not requested, then the single-quoted representation (for | |
3014 | * characters dumped as symbolic names) and integer code in | |
3015 | * decimal and hex are printed as a comment. | |
3016 | */ | |
3017 | ||
e63124bc MW |
3018 | static void dump_char(const union tvec_regval *rv, |
3019 | const struct tvec_regdef *rd, | |
3020 | unsigned style, | |
3021 | const struct gprintf_ops *gops, void *go) | |
3022 | { | |
67b5031e MW |
3023 | const char *p; |
3024 | unsigned f = 0; | |
3025 | #define f_semi 1u | |
3026 | ||
5c0f2e08 MW |
3027 | if (style&TVSF_RAW) { |
3028 | /* Print the raw character unconditionally in single quotes. */ | |
67b5031e | 3029 | |
5c0f2e08 MW |
3030 | gprintf(gops, go, "char:'"); |
3031 | format_char(gops, go, rv->i); | |
3032 | gprintf(gops, go, "'"); | |
3033 | } else { | |
3034 | /* Print ina pleasant human-readable way. */ | |
3035 | ||
3036 | /* Print a character name if we can find one. */ | |
3037 | p = find_charname(rv->i, (style&TVSF_COMPACT) ? CTF_SHORT : CTF_PREFER); | |
3038 | if (p) { | |
3039 | gprintf(gops, go, "%s", p); | |
3040 | if (style&TVSF_COMPACT) return; | |
3041 | else { gprintf(gops, go, " ;"); f |= f_semi; } | |
67b5031e | 3042 | } |
e63124bc | 3043 | |
5c0f2e08 MW |
3044 | /* If the character isn't @EOF@ then print it as a single-quoted thing. |
3045 | * In compact style, see if we can omit the quotes. | |
3046 | */ | |
3047 | if (rv->i >= 0) { | |
3048 | if (f&f_semi) gprintf(gops, go, " = "); | |
3049 | switch (rv->i) { | |
3050 | case ' ': case '\\': case '\'': quote: | |
3051 | format_char(gops, go, rv->i); | |
3052 | break; | |
3053 | default: | |
3054 | if (!(style&TVSF_COMPACT) || !isprint(rv->i)) goto quote; | |
3055 | gprintf(gops, go, "%c", (int)rv->i); | |
3056 | return; | |
3057 | } | |
3058 | } | |
3059 | ||
3060 | /* And the character code as an integer. */ | |
3061 | if (!(style&TVSF_COMPACT)) { | |
3062 | if (!(f&f_semi)) gprintf(gops, go, " ;"); | |
3063 | gprintf(gops, go, " = %ld = ", rv->i); | |
3064 | format_signed_hex(gops, go, rv->i); | |
3065 | } | |
e63124bc | 3066 | } |
67b5031e MW |
3067 | |
3068 | #undef f_semi | |
e63124bc MW |
3069 | } |
3070 | ||
c81c35df | 3071 | /* Character type definition. */ |
e63124bc | 3072 | const struct tvec_regty tvty_char = { |
3efcfd2d | 3073 | init_int, trivial_release, eq_int, |
e63124bc MW |
3074 | tobuf_char, frombuf_char, |
3075 | parse_char, dump_char | |
3076 | }; | |
3077 | ||
67b5031e MW |
3078 | /* --- @tvec_claimeq_char@ --- * |
3079 | * | |
3080 | * Arguments: @struct tvec_state *tv@ = test-vector state | |
3081 | * @int ch0, ch1@ = two character codes | |
3082 | * @const char *file@, @unsigned @lno@ = calling file and line | |
3083 | * @const char *expr@ = the expression to quote on failure | |
3084 | * | |
3085 | * Returns: Nonzero if @ch0@ and @ch1@ are equal, otherwise zero. | |
3086 | * | |
3087 | * Use: Check that values of @ch0@ and @ch1@ are equal. As for | |
3088 | * @tvec_claim@ above, a test case is automatically begun and | |
3089 | * ended if none is already underway. If the values are | |
3090 | * unequal, then @tvec_fail@ is called, quoting @expr@, and the | |
3091 | * mismatched values are dumped: @ch0@ is printed as the output | |
3092 | * value and @ch1@ is printed as the input reference. | |
3093 | */ | |
3094 | ||
e63124bc MW |
3095 | int tvec_claimeq_char(struct tvec_state *tv, int c0, int c1, |
3096 | const char *file, unsigned lno, const char *expr) | |
3097 | { | |
3efcfd2d | 3098 | tv->out[0].v.i = c0; tv->in[0].v.i = c1; |
e63124bc MW |
3099 | return (tvec_claimeq(tv, &tvty_char, 0, file, lno, expr)); |
3100 | } | |
3101 | ||
b64eb60f MW |
3102 | /*----- Text and byte strings ---------------------------------------------*/ |
3103 | ||
c81c35df MW |
3104 | /* --- @init_text@, @init_bytes@ --- * |
3105 | * | |
3106 | * Arguments: @union tvec_regval *rv@ = register value | |
3107 | * @const struct tvec_regdef *rd@ = register definition | |
3108 | * | |
3109 | * Returns: --- | |
3110 | * | |
3111 | * Use: Initialize a register value. | |
3112 | * | |
3113 | * Text and binary string values are initialized with a null | |
3114 | * pointer and zero length. | |
3115 | */ | |
3116 | ||
3117 | static void init_text(union tvec_regval *rv, const struct tvec_regdef *rd) | |
3118 | { rv->text.p = 0; rv->text.sz = 0; } | |
b64eb60f MW |
3119 | |
3120 | static void init_bytes(union tvec_regval *rv, const struct tvec_regdef *rd) | |
3121 | { rv->bytes.p = 0; rv->bytes.sz = 0; } | |
3122 | ||
c81c35df MW |
3123 | /* --- @release_string@, @release_bytes@ --- * |
3124 | * | |
3125 | * Arguments: @const union tvec_regval *rv@ = register value | |
3126 | * @const struct tvec_regdef *rd@ = register definition | |
3127 | * | |
3128 | * Returns: --- | |
3129 | * | |
3130 | * Use: Release resources held by a register value. | |
3131 | * | |
3132 | * Text and binary string buffers are freed. | |
3133 | */ | |
3134 | ||
3135 | static void release_text(union tvec_regval *rv, | |
3136 | const struct tvec_regdef *rd) | |
b1a20bee | 3137 | { free(rv->text.p); } |
b64eb60f MW |
3138 | |
3139 | static void release_bytes(union tvec_regval *rv, | |
3140 | const struct tvec_regdef *rd) | |
b1a20bee | 3141 | { free(rv->bytes.p); } |
b64eb60f | 3142 | |
c81c35df MW |
3143 | /* --- @eq_text@, @eq_bytes@ --- * |
3144 | * | |
3145 | * Arguments: @const union tvec_regval *rv0, *rv1@ = register values | |
3146 | * @const struct tvec_regdef *rd@ = register definition | |
3147 | * | |
3148 | * Returns: Nonzero if the values are equal, zero if unequal | |
3149 | * | |
3150 | * Use: Compare register values for equality. | |
3151 | */ | |
3152 | ||
3153 | static int eq_text(const union tvec_regval *rv0, | |
3154 | const union tvec_regval *rv1, | |
3155 | const struct tvec_regdef *rd) | |
b64eb60f | 3156 | { |
c81c35df MW |
3157 | return (rv0->text.sz == rv1->text.sz && |
3158 | (!rv0->text.sz || | |
3159 | MEMCMP(rv0->text.p, ==, rv1->text.p, rv1->text.sz))); | |
b64eb60f MW |
3160 | } |
3161 | ||
3162 | static int eq_bytes(const union tvec_regval *rv0, | |
3163 | const union tvec_regval *rv1, | |
3164 | const struct tvec_regdef *rd) | |
3165 | { | |
3166 | return (rv0->bytes.sz == rv1->bytes.sz && | |
3167 | (!rv0->bytes.sz || | |
3168 | MEMCMP(rv0->bytes.p, ==, rv1->bytes.p, rv1->bytes.sz))); | |
3169 | } | |
3170 | ||
c81c35df MW |
3171 | /* --- @tobuf_text@, @tobuf_bytes@ --- * |
3172 | * | |
3173 | * Arguments: @buf *b@ = buffer | |
3174 | * @const union tvec_regval *rv@ = register value | |
3175 | * @const struct tvec_regdef *rd@ = register definition | |
3176 | * | |
3177 | * Returns: Zero on success, %$-1$% on failure. | |
3178 | * | |
3179 | * Use: Serialize a register value to a buffer. | |
3180 | * | |
3181 | * Text and binary string values are serialized as a little- | |
3182 | * endian 64-bit length %$n$% in bytes followed by %$n$% bytes | |
3183 | * of string data. | |
3184 | */ | |
3185 | ||
3186 | static int tobuf_text(buf *b, const union tvec_regval *rv, | |
3187 | const struct tvec_regdef *rd) | |
3188 | { return (buf_putmem64l(b, rv->text.p, rv->text.sz)); } | |
b64eb60f MW |
3189 | |
3190 | static int tobuf_bytes(buf *b, const union tvec_regval *rv, | |
3191 | const struct tvec_regdef *rd) | |
c81c35df | 3192 | { return (buf_putmem64l(b, rv->bytes.p, rv->bytes.sz)); } |
b64eb60f | 3193 | |
c81c35df MW |
3194 | /* --- @frombuf_text@, @frombuf_bytes@ --- * |
3195 | * | |
3196 | * Arguments: @buf *b@ = buffer | |
3197 | * @union tvec_regval *rv@ = register value | |
3198 | * @const struct tvec_regdef *rd@ = register definition | |
3199 | * | |
3200 | * Returns: Zero on success, %$-1$% on failure. | |
3201 | * | |
3202 | * Use: Deserialize a register value from a buffer. | |
3203 | * | |
3204 | * Text and binary string values are serialized as a little- | |
3205 | * endian 64-bit length %$n$% in bytes followed by %$n$% bytes | |
3206 | * of string data. | |
3207 | */ | |
3208 | ||
3209 | static int frombuf_text(buf *b, union tvec_regval *rv, | |
3210 | const struct tvec_regdef *rd) | |
b64eb60f MW |
3211 | { |
3212 | const void *p; | |
3213 | size_t sz; | |
3214 | ||
c81c35df MW |
3215 | p = buf_getmem64l(b, &sz); if (!p) return (-1); |
3216 | tvec_alloctext(rv, sz); memcpy(rv->text.p, p, sz); rv->text.p[sz] = 0; | |
b64eb60f MW |
3217 | return (0); |
3218 | } | |
3219 | ||
3220 | static int frombuf_bytes(buf *b, union tvec_regval *rv, | |
3221 | const struct tvec_regdef *rd) | |
3222 | { | |
3223 | const void *p; | |
3224 | size_t sz; | |
3225 | ||
c81c35df | 3226 | p = buf_getmem64l(b, &sz); if (!p) return (-1); |
b64eb60f MW |
3227 | tvec_allocbytes(rv, sz); memcpy(rv->bytes.p, p, sz); |
3228 | return (0); | |
3229 | } | |
3230 | ||
c81c35df MW |
3231 | /* --- @check_string_length@ --- * |
3232 | * | |
3233 | * Arguments: @size_t sz@ = found string length | |
3234 | * @const struct tvec_urange *ur@ = acceptable range | |
3235 | * @struct tvec_state *tv@ = test-vector state | |
3236 | * | |
3237 | * Returns: Zero on success, %$-1$% on error. | |
3238 | * | |
3239 | * Use: Checks that @sz@ is within the bounds described by @ur@, | |
3240 | * reporting an error if not. | |
3241 | */ | |
3242 | ||
882a39c1 MW |
3243 | static int check_string_length(size_t sz, const struct tvec_urange *ur, |
3244 | struct tvec_state *tv) | |
b64eb60f | 3245 | { |
b1a20bee MW |
3246 | unsigned long uu; |
3247 | ||
3248 | if (ur) { | |
3249 | if (ur->min > sz || sz > ur->max) { | |
3250 | tvec_error(tv, "invalid string length %lu; must be in [%lu .. %lu]", | |
3251 | (unsigned long)sz, ur->min, ur->max); | |
3252 | return (-1); | |
3253 | } | |
3254 | if (ur->m && ur->m != 1) { | |
3255 | uu = sz%ur->m; | |
3256 | if (uu != ur->a%ur->m) { | |
3257 | tvec_error(tv, "invalid string length %lu == %lu =/= %lu (mod %lu)", | |
3258 | (unsigned long)sz, uu, ur->a, ur->m); | |
3259 | return (-1); | |
3260 | } | |
3261 | } | |
3262 | } | |
882a39c1 | 3263 | return (0); |
b64eb60f MW |
3264 | } |
3265 | ||
c81c35df MW |
3266 | /* --- @parse_text@, @parse_bytes@ --- * |
3267 | * | |
3268 | * Arguments: @union tvec_regval *rv@ = register value | |
3269 | * @const struct tvec_regdef *rd@ = register definition | |
3270 | * @struct tvec_state *tv@ = test-vector state | |
3271 | * | |
3272 | * Returns: Zero on success, %$-1$% on error. | |
3273 | * | |
3274 | * Use: Parse a register value from an input file. | |
3275 | * | |
3276 | * The input format for both kinds of strings is basically the | |
3277 | * same: a `compound string', consisting of | |
3278 | * | |
3279 | * * single-quoted strings, which are interpreted entirely | |
3280 | * literally, but can't contain single quotes or newlines; | |
3281 | * | |
3282 | * * double-quoted strings, in which `%|\|%'-escapes are | |
3283 | * interpreted as for characters; | |
3284 | * | |
3285 | * * character names, marked by an initial `%|#|%' sign; | |
3286 | * | |
3287 | * * special tokens marked by an initial `%|!|%' sign; or | |
3288 | * | |
3289 | * * barewords interpreted according to the current coding | |
3290 | * scheme. | |
3291 | * | |
3292 | * The special tokens are | |
3293 | * | |
3294 | * * `%|!bare|%', which causes subsequent sequences of | |
3295 | * barewords to be treated as plain text; | |
3296 | * | |
3297 | * * `%|!hex|%', `%|!base32|%', `%|!base64|%', which cause | |
3298 | * subsequent barewords to be decoded in the requested | |
3299 | * manner. | |
3300 | * | |
3301 | * * `%|!repeat|% %$n$% %|{|% %%\textit{string}%% %|}|%', | |
3302 | * which includes %$n$% copies of the (compound) string. | |
3303 | * | |
3304 | * The only difference between text and binary strings is that | |
3305 | * the initial coding scheme is %|bare|% for text strings and | |
3306 | * %|hex|% for binary strings. | |
3307 | */ | |
3308 | ||
3309 | static int parse_text(union tvec_regval *rv, const struct tvec_regdef *rd, | |
3310 | struct tvec_state *tv) | |
b64eb60f | 3311 | { |
c81c35df | 3312 | void *p = rv->text.p; |
b64eb60f | 3313 | |
c81c35df | 3314 | if (read_compound_string(&p, &rv->text.sz, TVCODE_BARE, 0, tv)) |
67b5031e | 3315 | return (-1); |
c81c35df MW |
3316 | rv->text.p = p; |
3317 | if (check_string_length(rv->text.sz, rd->arg.p, tv)) return (-1); | |
882a39c1 | 3318 | return (0); |
b64eb60f MW |
3319 | } |
3320 | ||
882a39c1 MW |
3321 | static int parse_bytes(union tvec_regval *rv, const struct tvec_regdef *rd, |
3322 | struct tvec_state *tv) | |
b64eb60f MW |
3323 | { |
3324 | void *p = rv->bytes.p; | |
3325 | ||
67b5031e MW |
3326 | if (read_compound_string(&p, &rv->bytes.sz, TVCODE_HEX, 0, tv)) |
3327 | return (-1); | |
882a39c1 MW |
3328 | rv->bytes.p = p; |
3329 | if (check_string_length(rv->bytes.sz, rd->arg.p, tv)) return (-1); | |
3330 | return (0); | |
b64eb60f MW |
3331 | } |
3332 | ||
c81c35df MW |
3333 | /* --- @dump_text@, @dump_bytes@ --- * |
3334 | * | |
3335 | * Arguments: @const union tvec_regval *rv@ = register value | |
3336 | * @const struct tvec_regdef *rd@ = register definition | |
3337 | * @unsigned style@ = output style (@TVSF_...@) | |
3338 | * @const struct gprintf_ops *gops@, @void *gp@ = format output | |
3339 | * | |
3340 | * Returns: --- | |
3341 | * | |
3342 | * Use: Dump a register value to the format output. | |
3343 | * | |
3344 | * Text string values are dumped as plain text, in double quotes | |
3345 | * if necessary, and using backslash escape sequences for | |
3346 | * nonprintable characters. Unless compact output is requested, | |
3347 | * strings consisting of multiple lines are dumped with each | |
3348 | * line of the string on a separate output line. | |
3349 | * | |
3350 | * Binary string values are dumped in hexadecimal. In compact | |
3351 | * style, the output simply consists of a single block of hex | |
3352 | * digits. Otherwise, the dump is a display consisting of | |
3353 | * groups of hex digits, with comments showing the offset (if | |
3354 | * the string is long enough) and the corresponding plain text. | |
3355 | * | |
5c0f2e08 | 3356 | * Empty strings are dumped as %|#empty|%. |
c81c35df MW |
3357 | */ |
3358 | ||
5c0f2e08 MW |
3359 | static void dump_empty(const char *ty, unsigned style, |
3360 | const struct gprintf_ops *gops, void *go) | |
3361 | { | |
3362 | if (style&TVSF_RAW) gprintf(gops, go, "%s:", ty); | |
3363 | if (!(style&TVSF_COMPACT)) gprintf(gops, go, "#empty"); | |
3364 | if (!(style&(TVSF_COMPACT | TVSF_RAW))) gprintf(gops, go, " ; = "); | |
3365 | if (!(style&TVSF_RAW)) gprintf(gops, go, "\"\""); | |
3366 | } | |
3367 | ||
3368 | ||
c81c35df MW |
3369 | static void dump_text(const union tvec_regval *rv, |
3370 | const struct tvec_regdef *rd, | |
3371 | unsigned style, | |
3372 | const struct gprintf_ops *gops, void *go) | |
b64eb60f MW |
3373 | { |
3374 | const unsigned char *p, *q, *l; | |
b64eb60f MW |
3375 | unsigned f = 0; |
3376 | #define f_nonword 1u | |
3377 | #define f_newline 2u | |
3378 | ||
5c0f2e08 | 3379 | if (!rv->text.sz) { dump_empty("text", style, gops, go); return; } |
b64eb60f | 3380 | |
c81c35df | 3381 | p = (const unsigned char *)rv->text.p; l = p + rv->text.sz; |
5c0f2e08 MW |
3382 | if (style&TVSF_RAW) { gprintf(gops, go, "text:"); goto quote; } |
3383 | else if (style&TVSF_COMPACT) goto quote; | |
3384 | ||
67b5031e MW |
3385 | switch (*p) { |
3386 | case '!': case '#': case ';': case '"': case '\'': | |
3387 | case '(': case '{': case '[': case ']': case '}': case ')': | |
3388 | f |= f_nonword; break; | |
3389 | } | |
b64eb60f MW |
3390 | for (q = p; q < l; q++) |
3391 | if (*q == '\n' && q != l - 1) f |= f_newline; | |
5c0f2e08 | 3392 | else if (!*q || !ISGRAPH(*q) || *q == '\\') f |= f_nonword; |
e63124bc | 3393 | if (f&f_newline) { gprintf(gops, go, "\n\t"); goto quote; } |
b64eb60f | 3394 | else if (f&f_nonword) goto quote; |
67b5031e | 3395 | |
c81c35df | 3396 | gops->putm(go, (const char *)p, rv->text.sz); |
67b5031e | 3397 | return; |
b64eb60f MW |
3398 | |
3399 | quote: | |
e63124bc | 3400 | gprintf(gops, go, "\""); |
b64eb60f | 3401 | for (q = p; q < l; q++) |
5c0f2e08 | 3402 | if (!ISPRINT(*q) || *q == '"') { |
e63124bc | 3403 | if (p < q) gops->putm(go, (const char *)p, q - p); |
67b5031e | 3404 | if (*q != '\n' || (style&TVSF_COMPACT)) |
3efcfd2d | 3405 | format_charesc(gops, go, *q, FCF_BRACE); |
67b5031e MW |
3406 | else { |
3407 | if (q + 1 == l) { gprintf(gops, go, "\\n\""); return; } | |
3408 | else gprintf(gops, go, "\\n\"\n\t\""); | |
3409 | } | |
3410 | p = q + 1; | |
b64eb60f | 3411 | } |
e63124bc MW |
3412 | if (p < q) gops->putm(go, (const char *)p, q - p); |
3413 | gprintf(gops, go, "\""); | |
b64eb60f MW |
3414 | |
3415 | #undef f_nonword | |
3416 | #undef f_newline | |
3417 | } | |
3418 | ||
3419 | static void dump_bytes(const union tvec_regval *rv, | |
3420 | const struct tvec_regdef *rd, | |
e63124bc MW |
3421 | unsigned style, |
3422 | const struct gprintf_ops *gops, void *go) | |
b64eb60f MW |
3423 | { |
3424 | const unsigned char *p = rv->bytes.p, *l = p + rv->bytes.sz; | |
3425 | size_t off, sz = rv->bytes.sz; | |
3426 | unsigned i, n; | |
3427 | int wd; | |
3428 | ||
5c0f2e08 | 3429 | if (!rv->text.sz) { dump_empty("bytes", style, gops, go); return; } |
b64eb60f | 3430 | |
5c0f2e08 | 3431 | if (style&(TVSF_COMPACT | TVSF_RAW)) { |
b1a20bee | 3432 | if (style&TVSF_RAW) gprintf(gops, go, "bytes:"); |
e63124bc | 3433 | while (p < l) gprintf(gops, go, "%02x", *p++); |
b64eb60f MW |
3434 | return; |
3435 | } | |
3436 | ||
e63124bc | 3437 | if (sz > 16) gprintf(gops, go, "\n\t"); |
b64eb60f MW |
3438 | |
3439 | off = 0; wd = hex_width(sz); | |
3440 | while (p < l) { | |
3441 | if (l - p < 16) n = l - p; | |
3442 | else n = 16; | |
3443 | ||
67b5031e | 3444 | for (i = 0; i < n; i++) { |
e63124bc MW |
3445 | if (i < n) gprintf(gops, go, "%02x", p[i]); |
3446 | else gprintf(gops, go, " "); | |
67b5031e | 3447 | if (i < n - 1 && i%4 == 3) gprintf(gops, go, " "); |
b64eb60f | 3448 | } |
e63124bc MW |
3449 | gprintf(gops, go, " ; "); |
3450 | if (sz > 16) gprintf(gops, go, "[%0*lx] ", wd, (unsigned long)off); | |
b64eb60f | 3451 | for (i = 0; i < n; i++) |
e63124bc | 3452 | gprintf(gops, go, "%c", isprint(p[i]) ? p[i] : '.'); |
b64eb60f | 3453 | p += n; off += n; |
e63124bc | 3454 | if (p < l) gprintf(gops, go, "\n\t"); |
b64eb60f MW |
3455 | } |
3456 | } | |
3457 | ||
c81c35df MW |
3458 | /* Text and byte string type definitions. */ |
3459 | const struct tvec_regty tvty_text = { | |
3460 | init_text, release_text, eq_text, | |
3461 | tobuf_text, frombuf_text, | |
3462 | parse_text, dump_text | |
b64eb60f | 3463 | }; |
b64eb60f | 3464 | const struct tvec_regty tvty_bytes = { |
e63124bc | 3465 | init_bytes, release_bytes, eq_bytes, |
b64eb60f MW |
3466 | tobuf_bytes, frombuf_bytes, |
3467 | parse_bytes, dump_bytes | |
3468 | }; | |
3469 | ||
c81c35df | 3470 | /* --- @tvec_claimeq_text@ --- * |
67b5031e MW |
3471 | * |
3472 | * Arguments: @struct tvec_state *tv@ = test-vector state | |
3473 | * @const char *p0@, @size_t sz0@ = first string with length | |
3474 | * @const char *p1@, @size_t sz1@ = second string with length | |
3475 | * @const char *file@, @unsigned @lno@ = calling file and line | |
3476 | * @const char *expr@ = the expression to quote on failure | |
3477 | * | |
3478 | * Returns: Nonzero if the strings at @p0@ and @p1@ are equal, otherwise | |
3479 | * zero. | |
3480 | * | |
3481 | * Use: Check that strings at @p0@ and @p1@ are equal. As for | |
3482 | * @tvec_claim@ above, a test case is automatically begun and | |
3483 | * ended if none is already underway. If the values are | |
3484 | * unequal, then @tvec_fail@ is called, quoting @expr@, and the | |
3485 | * mismatched values are dumped: @p0@ is printed as the output | |
3486 | * value and @p1@ is printed as the input reference. | |
3487 | */ | |
3488 | ||
c81c35df MW |
3489 | int tvec_claimeq_text(struct tvec_state *tv, |
3490 | const char *p0, size_t sz0, | |
3491 | const char *p1, size_t sz1, | |
3492 | const char *file, unsigned lno, const char *expr) | |
b64eb60f | 3493 | { |
b1a20bee MW |
3494 | tv->out[0].v.text.p = UNCONST(char, p0); tv->out[0].v.text.sz = sz0; |
3495 | tv->in[0].v.text.p =UNCONST(char, p1); tv->in[0].v.text.sz = sz1; | |
c81c35df | 3496 | return (tvec_claimeq(tv, &tvty_text, 0, file, lno, expr)); |
b64eb60f MW |
3497 | } |
3498 | ||
c81c35df | 3499 | /* --- @tvec_claimeq_textz@ --- * |
67b5031e MW |
3500 | * |
3501 | * Arguments: @struct tvec_state *tv@ = test-vector state | |
3502 | * @const char *p0, *p1@ = two strings to compare | |
3503 | * @const char *file@, @unsigned @lno@ = calling file and line | |
3504 | * @const char *expr@ = the expression to quote on failure | |
3505 | * | |
3506 | * Returns: Nonzero if the strings at @p0@ and @p1@ are equal, otherwise | |
3507 | * zero. | |
3508 | * | |
3509 | * Use: Check that strings at @p0@ and @p1@ are equal, as for | |
3510 | * @tvec_claimeq_string@, except that the strings are assumed | |
3511 | * null-terminated, so their lengths don't need to be supplied | |
3512 | * explicitly. | |
3513 | */ | |
3514 | ||
c81c35df MW |
3515 | int tvec_claimeq_textz(struct tvec_state *tv, |
3516 | const char *p0, const char *p1, | |
3517 | const char *file, unsigned lno, const char *expr) | |
b64eb60f | 3518 | { |
b1a20bee MW |
3519 | tv->out[0].v.text.p = UNCONST(char, p0); tv->out[0].v.text.sz = strlen(p0); |
3520 | tv->in[0].v.text.p = UNCONST(char, p1); tv->in[0].v.text.sz = strlen(p1); | |
c81c35df | 3521 | return (tvec_claimeq(tv, &tvty_text, 0, file, lno, expr)); |
b64eb60f MW |
3522 | } |
3523 | ||
67b5031e MW |
3524 | /* --- @tvec_claimeq_bytes@ --- * |
3525 | * | |
3526 | * Arguments: @struct tvec_state *tv@ = test-vector state | |
3527 | * @const void *p0@, @size_t sz0@ = first string with length | |
3528 | * @const void *p1@, @size_t sz1@ = second string with length | |
3529 | * @const char *file@, @unsigned @lno@ = calling file and line | |
3530 | * @const char *expr@ = the expression to quote on failure | |
3531 | * | |
3532 | * Returns: Nonzero if the strings at @p0@ and @p1@ are equal, otherwise | |
3533 | * zero. | |
3534 | * | |
3535 | * Use: Check that binary strings at @p0@ and @p1@ are equal. As for | |
3536 | * @tvec_claim@ above, a test case is automatically begun and | |
3537 | * ended if none is already underway. If the values are | |
3538 | * unequal, then @tvec_fail@ is called, quoting @expr@, and the | |
3539 | * mismatched values are dumped: @p0@ is printed as the output | |
3540 | * value and @p1@ is printed as the input reference. | |
3541 | */ | |
3542 | ||
b64eb60f MW |
3543 | int tvec_claimeq_bytes(struct tvec_state *tv, |
3544 | const void *p0, size_t sz0, | |
3545 | const void *p1, size_t sz1, | |
3546 | const char *file, unsigned lno, const char *expr) | |
3547 | { | |
b1a20bee MW |
3548 | tv->out[0].v.bytes.p = UNCONST(void, p0); tv->out[0].v.bytes.sz = sz0; |
3549 | tv->in[0].v.bytes.p = UNCONST(void, p1); tv->in[0].v.bytes.sz = sz1; | |
b64eb60f MW |
3550 | return (tvec_claimeq(tv, &tvty_bytes, 0, file, lno, expr)); |
3551 | } | |
3552 | ||
c81c35df | 3553 | /* --- @tvec_alloctext@, @tvec_allocbytes@ --- * |
67b5031e MW |
3554 | * |
3555 | * Arguments: @union tvec_regval *rv@ = register value | |
3556 | * @size_t sz@ = required size | |
3557 | * | |
3558 | * Returns: --- | |
3559 | * | |
3560 | * Use: Allocated space in a text or binary string register. If the | |
3561 | * current register size is sufficient, its buffer is left | |
3562 | * alone; otherwise, the old buffer, if any, is freed and a | |
3563 | * fresh buffer allocated. These functions are not intended to | |
3564 | * be used to adjust a buffer repeatedly, e.g., while building | |
3565 | * output incrementally: (a) they will perform badly, and (b) | |
3566 | * the old buffer contents are simply discarded if reallocation | |
3567 | * is necessary. Instead, use a @dbuf@ or @dstr@. | |
3568 | * | |
c81c35df | 3569 | * The @tvec_alloctext@ function sneakily allocates an extra |
67b5031e MW |
3570 | * byte for a terminating zero. The @tvec_allocbytes@ function |
3571 | * doesn't do this. | |
3572 | */ | |
3573 | ||
c81c35df | 3574 | void tvec_alloctext(union tvec_regval *rv, size_t sz) |
67b5031e | 3575 | { |
b1a20bee MW |
3576 | if (rv->text.sz <= sz) |
3577 | { free(rv->text.p); rv->text.p = x_alloc(&arena_stdlib, sz + 1); } | |
3578 | memset(rv->text.p, '?', sz); rv->text.sz = sz; | |
67b5031e MW |
3579 | } |
3580 | ||
3581 | void tvec_allocbytes(union tvec_regval *rv, size_t sz) | |
3582 | { | |
b1a20bee MW |
3583 | if (rv->bytes.sz < sz) |
3584 | { free(rv->bytes.p); rv->bytes.p = x_alloc(&arena_stdlib, sz); } | |
3585 | memset(rv->bytes.p, '?', sz); rv->bytes.sz = sz; | |
67b5031e MW |
3586 | } |
3587 | ||
b64eb60f MW |
3588 | /*----- Buffer type -------------------------------------------------------*/ |
3589 | ||
adec5584 MW |
3590 | /* --- @init_buffer@ --- * |
3591 | * | |
3592 | * Arguments: @union tvec_regval *rv@ = register value | |
3593 | * @const struct tvec_regdef *rd@ = register definition | |
3594 | * | |
3595 | * Returns: --- | |
3596 | * | |
3597 | * Use: Initialize a register value. | |
3598 | * | |
3599 | * Buffer values values are initialized with a null pointer, | |
3600 | * zero length, and zero residue, modulus, and offset. | |
3601 | */ | |
3602 | ||
3603 | static void init_buffer(union tvec_regval *rv, const struct tvec_regdef *rd) | |
3604 | { rv->buf.p = 0; rv->buf.sz = rv->buf.a = rv->buf.m = rv->buf.off = 0; } | |
3605 | ||
3606 | /* --- @release_buffer@, @release_bytes@ --- * | |
3607 | * | |
3608 | * Arguments: @const union tvec_regval *rv@ = register value | |
3609 | * @const struct tvec_regdef *rd@ = register definition | |
3610 | * | |
3611 | * Returns: --- | |
3612 | * | |
3613 | * Use: Release resources held by a register value. | |
3614 | * | |
3615 | * Buffers are freed. | |
3616 | */ | |
3617 | ||
3618 | static void release_buffer(union tvec_regval *rv, | |
3619 | const struct tvec_regdef *rd) | |
b1a20bee | 3620 | { if (rv->buf.p) free(rv->buf.p - rv->buf.off); } |
c81c35df MW |
3621 | |
3622 | /* --- @eq_buffer@ --- * | |
3623 | * | |
3624 | * Arguments: @const union tvec_regval *rv0, *rv1@ = register values | |
3625 | * @const struct tvec_regdef *rd@ = register definition | |
3626 | * | |
3627 | * Returns: Nonzero if the values are equal, zero if unequal | |
3628 | * | |
3629 | * Use: Compare register values for equality. | |
3630 | * | |
adec5584 MW |
3631 | * Buffer values are equal if and only if their sizes and |
3632 | * alignment parameters are equal; their contents are | |
3633 | * %%\emph{not}%% compared. | |
c81c35df MW |
3634 | */ |
3635 | ||
b64eb60f MW |
3636 | static int eq_buffer(const union tvec_regval *rv0, |
3637 | const union tvec_regval *rv1, | |
3638 | const struct tvec_regdef *rd) | |
adec5584 MW |
3639 | { |
3640 | return (rv0->buf.sz == rv1->buf.sz && | |
3641 | rv0->buf.a == rv1->buf.a && | |
3642 | rv0->buf.m == rv1->buf.m); | |
3643 | } | |
b64eb60f | 3644 | |
c81c35df MW |
3645 | /* --- @tobuf_buffer@ --- * |
3646 | * | |
3647 | * Arguments: @buf *b@ = buffer | |
3648 | * @const union tvec_regval *rv@ = register value | |
3649 | * @const struct tvec_regdef *rd@ = register definition | |
3650 | * | |
3651 | * Returns: Zero on success, %$-1$% on failure. | |
3652 | * | |
3653 | * Use: Serialize a register value to a buffer. | |
3654 | * | |
adec5584 MW |
3655 | * Buffer values are serialized as their lengths, residues, and |
3656 | * moduli, as unsigned integers. | |
c81c35df MW |
3657 | */ |
3658 | ||
b64eb60f MW |
3659 | static int tobuf_buffer(buf *b, const union tvec_regval *rv, |
3660 | const struct tvec_regdef *rd) | |
adec5584 MW |
3661 | { |
3662 | return (unsigned_to_buf(b, rv->buf.sz) || | |
3663 | unsigned_to_buf(b, rv->buf.a) || | |
3664 | unsigned_to_buf(b, rv->buf.m)); | |
3665 | } | |
c81c35df MW |
3666 | |
3667 | /* --- @frombuf_buffer@ --- * | |
3668 | * | |
3669 | * Arguments: @buf *b@ = buffer | |
3670 | * @union tvec_regval *rv@ = register value | |
3671 | * @const struct tvec_regdef *rd@ = register definition | |
3672 | * | |
3673 | * Returns: Zero on success, %$-1$% on failure. | |
3674 | * | |
3675 | * Use: Deserialize a register value from a buffer. | |
3676 | * | |
3677 | * Buffer values are serialized as just their lengths, as | |
3678 | * unsigned integers. The buffer is allocated on | |
3679 | * deserialization and filled with a distinctive pattern. | |
3680 | */ | |
3681 | ||
b64eb60f MW |
3682 | static int frombuf_buffer(buf *b, union tvec_regval *rv, |
3683 | const struct tvec_regdef *rd) | |
3684 | { | |
adec5584 | 3685 | unsigned long sz, a, m; |
b64eb60f | 3686 | |
adec5584 MW |
3687 | if (unsigned_from_buf(b, &sz)) return (-1); |
3688 | if (unsigned_from_buf(b, &a)) return (-1); | |
3689 | if (unsigned_from_buf(b, &m)) return (-1); | |
3690 | if (sz > (size_t)-1 || a > (size_t)-1 || m > (size_t)-1) | |
3691 | { buf_break(b); return (-1); } | |
3692 | rv->buf.sz = sz; rv->buf.a = a; rv->buf.m = m; | |
b64eb60f MW |
3693 | return (0); |
3694 | } | |
3695 | ||
c81c35df MW |
3696 | /* --- @parse_buffer@ --- * |
3697 | * | |
3698 | * Arguments: @union tvec_regval *rv@ = register value | |
3699 | * @const struct tvec_regdef *rd@ = register definition | |
3700 | * @struct tvec_state *tv@ = test-vector state | |
3701 | * | |
3702 | * Returns: Zero on success, %$-1$% on error. | |
3703 | * | |
3704 | * Use: Parse a register value from an input file. | |
3705 | * | |
c4ccbbf9 MW |
3706 | * The input format for a buffer value is a size, followed by an |
3707 | * optional `%|@$%' and an alignment quantum and a further | |
3708 | * optional `%|+|%' and an alignment offset. The size, quantum, | |
3709 | * and offset are syntactically sizes. | |
c81c35df | 3710 | * |
c4ccbbf9 | 3711 | * The buffer is not allocated. |
c81c35df MW |
3712 | */ |
3713 | ||
882a39c1 MW |
3714 | static int parse_buffer(union tvec_regval *rv, |
3715 | const struct tvec_regdef *rd, | |
3716 | struct tvec_state *tv) | |
b64eb60f | 3717 | { |
c4ccbbf9 | 3718 | unsigned long sz, a = 0, m = 0; |
adec5584 | 3719 | int ch, rc; |
b64eb60f | 3720 | |
c4ccbbf9 MW |
3721 | if (parse_szint(tv, &sz, "@;", "buffer length")) { rc = -1; goto end; } |
3722 | if (check_unsigned_range(sz, &tvrange_size, tv, "buffer length")) | |
3723 | { rc = -1; goto end; } | |
adec5584 | 3724 | if (check_string_length(sz, rd->arg.p, tv)) { rc = -1; goto end; } |
b64eb60f | 3725 | |
adec5584 MW |
3726 | tvec_skipspc(tv); |
3727 | ch = getc(tv->fp); | |
3728 | if (ch == ';' || ch == '\n') { ungetc(ch, tv->fp); goto done; } | |
3729 | else if (ch != '@') { rc = tvec_syntax(tv, ch, "`@'"); goto end; } | |
3730 | ||
c4ccbbf9 MW |
3731 | if (parse_szint(tv, &m, "+;", "alignment quantum")) { rc = -1; goto end; } |
3732 | if (check_unsigned_range(a, &tvrange_size, tv, "alignment quantum")) | |
3733 | { rc = -1; goto end; } | |
adec5584 MW |
3734 | if (m == 1) m = 0; |
3735 | ||
3736 | tvec_skipspc(tv); | |
3737 | ch = getc(tv->fp); | |
3738 | if (ch == ';' || ch == '\n') { ungetc(ch, tv->fp); goto done; } | |
3739 | else if (ch != '+') { rc = tvec_syntax(tv, ch, "`+'"); goto end; } | |
3740 | ||
c4ccbbf9 MW |
3741 | if (parse_szint(tv, &a, ";", "alignment offset")) { rc = -1; goto end; } |
3742 | if (check_unsigned_range(m, &tvrange_size, tv, "alignment offset")) | |
3743 | { rc = -1; goto end; } | |
adec5584 MW |
3744 | if (a >= m) { |
3745 | rc = tvec_error(tv, "alignment offset %lu >= quantum %lu", | |
3746 | (unsigned long)a, (unsigned long)m); | |
3747 | goto end; | |
b64eb60f | 3748 | } |
b64eb60f | 3749 | |
adec5584 | 3750 | done: |
882a39c1 | 3751 | if (tvec_flushtoeol(tv, 0)) { rc = -1; goto end; } |
adec5584 | 3752 | rv->buf.sz = sz; rv->buf.a = a; rv->buf.m = m; |
882a39c1 MW |
3753 | rc = 0; |
3754 | end: | |
adec5584 | 3755 | return (rc); |
b64eb60f MW |
3756 | } |
3757 | ||
c81c35df MW |
3758 | /* --- @dump_buffer@ --- * |
3759 | * | |
3760 | * Arguments: @const union tvec_regval *rv@ = register value | |
3761 | * @const struct tvec_regdef *rd@ = register definition | |
3762 | * @unsigned style@ = output style (@TVSF_...@) | |
3763 | * @const struct gprintf_ops *gops@, @void *gp@ = format output | |
3764 | * | |
3765 | * Returns: --- | |
3766 | * | |
3767 | * Use: Dump a register value to the format output. | |
3768 | * | |
c4ccbbf9 MW |
3769 | * Buffer values are dumped as their size, with the alignment |
3770 | * quantum and alignment offset if these are non-default. | |
c81c35df MW |
3771 | */ |
3772 | ||
b64eb60f MW |
3773 | static void dump_buffer(const union tvec_regval *rv, |
3774 | const struct tvec_regdef *rd, | |
e63124bc MW |
3775 | unsigned style, |
3776 | const struct gprintf_ops *gops, void *go) | |
b64eb60f | 3777 | { |
b1a20bee | 3778 | if (style&TVSF_RAW) gprintf(gops, go, "buffer:"); |
adec5584 MW |
3779 | format_size(gops, go, rv->buf.sz, style); |
3780 | if (rv->buf.m) { | |
5c0f2e08 | 3781 | gprintf(gops, go, style&(TVSF_COMPACT | TVSF_RAW) ? "@" : " @ "); |
adec5584 MW |
3782 | format_size(gops, go, rv->buf.m, style); |
3783 | if (rv->buf.a) { | |
5c0f2e08 | 3784 | gprintf(gops, go, style&(TVSF_COMPACT | TVSF_RAW) ? "+" : " + "); |
adec5584 MW |
3785 | format_size(gops, go, rv->buf.a, style); |
3786 | } | |
3787 | } | |
b1a20bee | 3788 | if (!(style&(TVSF_COMPACT | TVSF_RAW))) { |
13ee7406 | 3789 | gprintf(gops, go, " ; = %lu", (unsigned long)rv->buf.sz); |
adec5584 | 3790 | if (rv->buf.m) { |
13ee7406 MW |
3791 | gprintf(gops, go, " @ %lu", (unsigned long)rv->buf.m); |
3792 | if (rv->buf.a) gprintf(gops, go, " + %lu", (unsigned long)rv->buf.a); | |
adec5584 MW |
3793 | } |
3794 | gprintf(gops, go, " = "); format_unsigned_hex(gops, go, rv->buf.sz); | |
3795 | if (rv->buf.m) { | |
3796 | gprintf(gops, go, " @ "); format_unsigned_hex(gops, go, rv->buf.m); | |
3797 | if (rv->buf.a) { | |
3798 | gprintf(gops, go, " + "); | |
3799 | format_unsigned_hex(gops, go, rv->buf.a); | |
3800 | } | |
3801 | } | |
b64eb60f MW |
3802 | } |
3803 | } | |
3804 | ||
c81c35df | 3805 | /* Buffer type definition. */ |
b64eb60f | 3806 | const struct tvec_regty tvty_buffer = { |
adec5584 | 3807 | init_buffer, release_buffer, eq_buffer, |
b64eb60f MW |
3808 | tobuf_buffer, frombuf_buffer, |
3809 | parse_buffer, dump_buffer | |
3810 | }; | |
3811 | ||
adec5584 MW |
3812 | /* --- @tvec_initbuffer@ --- * |
3813 | * | |
3814 | * Arguments: @union tvec_regval *rv@ = register value | |
d056fbdf | 3815 | * @const union tvec_regval *ref@ = source buffer |
adec5584 MW |
3816 | * @size_t sz@ = size to allocate |
3817 | * | |
3818 | * Returns: --- | |
3819 | * | |
d056fbdf | 3820 | * Use: Initialize the alignment parameters in @rv@ to match @ref@, |
adec5584 MW |
3821 | * and the size to @sz@. |
3822 | */ | |
3823 | ||
3824 | void tvec_initbuffer(union tvec_regval *rv, | |
d056fbdf MW |
3825 | const union tvec_regval *ref, size_t sz) |
3826 | { rv->buf.sz = sz; rv->buf.a = ref->buf.a; rv->buf.m = ref->buf.m; } | |
adec5584 MW |
3827 | |
3828 | /* --- @tvec_allocbuffer@ --- * | |
3829 | * | |
3830 | * Arguments: @union tvec_regval *rv@ = register value | |
3831 | * | |
3832 | * Returns: --- | |
3833 | * | |
3834 | * Use: Allocate @sz@ bytes to the buffer and fill the space with a | |
3835 | * distinctive pattern. | |
3836 | */ | |
3837 | ||
3838 | void tvec_allocbuffer(union tvec_regval *rv) | |
3839 | { | |
3840 | unsigned char *p; size_t n; | |
3841 | ||
b1a20bee | 3842 | if (rv->buf.p) free(rv->buf.p - rv->buf.off); |
adec5584 MW |
3843 | |
3844 | if (rv->buf.m < 2) { | |
b1a20bee | 3845 | rv->buf.p = x_alloc(&arena_stdlib, rv->buf.sz); rv->buf.off = 0; |
adec5584 | 3846 | } else { |
b1a20bee | 3847 | p = x_alloc(&arena_stdlib, rv->buf.sz + rv->buf.m - 1); |
adec5584 MW |
3848 | n = (size_t)p%rv->buf.m; |
3849 | rv->buf.off = (rv->buf.a - n + rv->buf.m)%rv->buf.m; | |
3850 | rv->buf.p = p + rv->buf.off; | |
3851 | } | |
3852 | memset(rv->buf.p, '?', rv->buf.sz); | |
3853 | } | |
3854 | ||
b64eb60f | 3855 | /*----- That's all, folks -------------------------------------------------*/ |