1/* 2 * Copyright (c) 2002 - 2005 Tony Finch <dot@dotat.at>. All rights reserved. 3 * 4 * This code is derived from software contributed to Berkeley by Dave Yost. 5 * It was rewritten to support ANSI C by Tony Finch. The original version of 6 * unifdef carried the following copyright notice. None of its code remains 7 * in this version (though some of the names remain). 8 * 9 * Copyright (c) 1985, 1993 10 * The Regents of the University of California. All rights reserved. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 */ 33 34#include <sys/cdefs.h> 35 36#ifndef lint 37#ifdef __IDSTRING 38__IDSTRING(Berkeley, "@(#)unifdef.c 8.1 (Berkeley) 6/6/93"); 39__IDSTRING(NetBSD, "$NetBSD: unifdef.c,v 1.8 2000/07/03 02:51:36 matt Exp $"); 40__IDSTRING(dotat, "$dotat: things/unifdef.c,v 1.171 2005/03/08 12:38:48 fanf2 Exp $"); 41#endif 42#endif /* not lint */ 43#ifdef __FBSDID 44__FBSDID("$FreeBSD: /repoman/r/ncvs/src/usr.bin/unifdef/unifdef.c,v 1.20 2005/05/21 09:55:09 ru Exp $"); 45#endif 46 47/* 48 * unifdef - remove ifdef'ed lines 49 * 50 * Wishlist: 51 * provide an option which will append the name of the 52 * appropriate symbol after #else's and #endif's 53 * provide an option which will check symbols after 54 * #else's and #endif's to see that they match their 55 * corresponding #ifdef or #ifndef 56 * 57 * The first two items above require better buffer handling, which would 58 * also make it possible to handle all "dodgy" directives correctly. 59 */ 60 61#include <ctype.h> 62#include <err.h> 63#include <stdarg.h> 64#include <stdbool.h> 65#include <stdio.h> 66#include <stdlib.h> 67#include <string.h> 68#include <unistd.h> 69 70size_t strlcpy(char *dst, const char *src, size_t siz); 71 72/* types of input lines: */ 73typedef enum { 74 LT_TRUEI, /* a true #if with ignore flag */ 75 LT_FALSEI, /* a false #if with ignore flag */ 76 LT_IF, /* an unknown #if */ 77 LT_TRUE, /* a true #if */ 78 LT_FALSE, /* a false #if */ 79 LT_ELIF, /* an unknown #elif */ 80 LT_ELTRUE, /* a true #elif */ 81 LT_ELFALSE, /* a false #elif */ 82 LT_ELSE, /* #else */ 83 LT_ENDIF, /* #endif */ 84 LT_DODGY, /* flag: directive is not on one line */ 85 LT_DODGY_LAST = LT_DODGY + LT_ENDIF, 86 LT_PLAIN, /* ordinary line */ 87 LT_EOF, /* end of file */ 88 LT_COUNT 89} Linetype; 90 91static char const * const linetype_name[] = { 92 "TRUEI", "FALSEI", "IF", "TRUE", "FALSE", 93 "ELIF", "ELTRUE", "ELFALSE", "ELSE", "ENDIF", 94 "DODGY TRUEI", "DODGY FALSEI", 95 "DODGY IF", "DODGY TRUE", "DODGY FALSE", 96 "DODGY ELIF", "DODGY ELTRUE", "DODGY ELFALSE", 97 "DODGY ELSE", "DODGY ENDIF", 98 "PLAIN", "EOF" 99}; 100 101/* state of #if processing */ 102typedef enum { 103 IS_OUTSIDE, 104 IS_FALSE_PREFIX, /* false #if followed by false #elifs */ 105 IS_TRUE_PREFIX, /* first non-false #(el)if is true */ 106 IS_PASS_MIDDLE, /* first non-false #(el)if is unknown */ 107 IS_FALSE_MIDDLE, /* a false #elif after a pass state */ 108 IS_TRUE_MIDDLE, /* a true #elif after a pass state */ 109 IS_PASS_ELSE, /* an else after a pass state */ 110 IS_FALSE_ELSE, /* an else after a true state */ 111 IS_TRUE_ELSE, /* an else after only false states */ 112 IS_FALSE_TRAILER, /* #elifs after a true are false */ 113 IS_COUNT 114} Ifstate; 115 116static char const * const ifstate_name[] = { 117 "OUTSIDE", "FALSE_PREFIX", "TRUE_PREFIX", 118 "PASS_MIDDLE", "FALSE_MIDDLE", "TRUE_MIDDLE", 119 "PASS_ELSE", "FALSE_ELSE", "TRUE_ELSE", 120 "FALSE_TRAILER" 121}; 122 123/* state of comment parser */ 124typedef enum { 125 NO_COMMENT = false, /* outside a comment */ 126 C_COMMENT, /* in a comment like this one */ 127 CXX_COMMENT, /* between // and end of line */ 128 STARTING_COMMENT, /* just after slash-backslash-newline */ 129 FINISHING_COMMENT, /* star-backslash-newline in a C comment */ 130 CHAR_LITERAL, /* inside '' */ 131 STRING_LITERAL /* inside "" */ 132} Comment_state; 133 134static char const * const comment_name[] = { 135 "NO", "C", "CXX", "STARTING", "FINISHING", "CHAR", "STRING" 136}; 137 138/* state of preprocessor line parser */ 139typedef enum { 140 LS_START, /* only space and comments on this line */ 141 LS_HASH, /* only space, comments, and a hash */ 142 LS_DIRTY /* this line can't be a preprocessor line */ 143} Line_state; 144 145static char const * const linestate_name[] = { 146 "START", "HASH", "DIRTY" 147}; 148 149/* 150 * Minimum translation limits from ISO/IEC 9899:1999 5.2.4.1 151 */ 152#define MAXDEPTH 64 /* maximum #if nesting */ 153#define MAXLINE 4096 /* maximum length of line */ 154#define MAXSYMS 4096 /* maximum number of symbols */ 155 156/* 157 * Sometimes when editing a keyword the replacement text is longer, so 158 * we leave some space at the end of the tline buffer to accommodate this. 159 */ 160#define EDITSLOP 10 161 162/* 163 * Globals. 164 */ 165 166static bool complement; /* -c: do the complement */ 167static bool debugging; /* -d: debugging reports */ 168static bool iocccok; /* -e: fewer IOCCC errors */ 169static bool killconsts; /* -k: eval constant #ifs */ 170static bool lnblank; /* -l: blank deleted lines */ 171static bool lnnum; /* -n: add #line directives */ 172static bool symlist; /* -s: output symbol list */ 173static bool text; /* -t: this is a text file */ 174 175static const char *symname[MAXSYMS]; /* symbol name */ 176static const char *value[MAXSYMS]; /* -Dsym=value */ 177static bool ignore[MAXSYMS]; /* -iDsym or -iUsym */ 178static int nsyms; /* number of symbols */ 179 180static FILE *input; /* input file pointer */ 181static const char *filename; /* input file name */ 182static int linenum; /* current line number */ 183 184static char tline[MAXLINE+EDITSLOP];/* input buffer plus space */ 185static char *keyword; /* used for editing #elif's */ 186 187static Comment_state incomment; /* comment parser state */ 188static Line_state linestate; /* #if line parser state */ 189static Ifstate ifstate[MAXDEPTH]; /* #if processor state */ 190static bool ignoring[MAXDEPTH]; /* ignore comments state */ 191static int stifline[MAXDEPTH]; /* start of current #if */ 192static int depth; /* current #if nesting */ 193static int delcount; /* count of deleted lines */ 194static bool keepthis; /* don't delete constant #if */ 195 196static int exitstat; /* program exit status */ 197 198static void addsym(bool, bool, char *); 199static void debug(const char *, ...); 200static void done(void); 201static void error(const char *); 202static int findsym(const char *); 203static void flushline(bool); 204static Linetype getline(void); 205static Linetype ifeval(const char **); 206static void ignoreoff(void); 207static void ignoreon(void); 208static void keywordedit(const char *); 209static void nest(void); 210static void process(void); 211static const char *skipcomment(const char *); 212static const char *skipsym(const char *); 213static void state(Ifstate); 214static int strlcmp(const char *, const char *, size_t); 215static void unnest(void); 216static void usage(void); 217 218#define endsym(c) (!isalpha((unsigned char)c) && !isdigit((unsigned char)c) && c != '_') 219 220/* 221 * The main program. 222 */ 223int 224main(int argc, char *argv[]) 225{ 226 int opt; 227 228 while ((opt = getopt(argc, argv, "i:D:U:I:cdeklnst")) != -1) 229 switch (opt) { 230 case 'i': /* treat stuff controlled by these symbols as text */ 231 /* 232 * For strict backwards-compatibility the U or D 233 * should be immediately after the -i but it doesn't 234 * matter much if we relax that requirement. 235 */ 236 opt = *optarg++; 237 if (opt == 'D') 238 addsym(true, true, optarg); 239 else if (opt == 'U') 240 addsym(true, false, optarg); 241 else 242 usage(); 243 break; 244 case 'D': /* define a symbol */ 245 addsym(false, true, optarg); 246 break; 247 case 'U': /* undef a symbol */ 248 addsym(false, false, optarg); 249 break; 250 case 'I': 251 /* no-op for compatibility with cpp */ 252 break; 253 case 'c': /* treat -D as -U and vice versa */ 254 complement = true; 255 break; 256 case 'd': 257 debugging = true; 258 break; 259 case 'e': /* fewer errors from dodgy lines */ 260 iocccok = true; 261 break; 262 case 'k': /* process constant #ifs */ 263 killconsts = true; 264 break; 265 case 'l': /* blank deleted lines instead of omitting them */ 266 lnblank = true; 267 break; 268 case 'n': /* add #line directive after deleted lines */ 269 lnnum = true; 270 break; 271 case 's': /* only output list of symbols that control #ifs */ 272 symlist = true; 273 break; 274 case 't': /* don't parse C comments */ 275 text = true; 276 break; 277 default: 278 usage(); 279 } 280 argc -= optind; 281 argv += optind; 282 if (argc > 1) { 283 errx(2, "can only do one file"); 284 } else if (argc == 1 && strcmp(*argv, "-") != 0) { 285 filename = *argv; 286 input = fopen(filename, "r"); 287 if (input == NULL) 288 err(2, "can't open %s", filename); 289 } else { 290 filename = "[stdin]"; 291 input = stdin; 292 } 293 process(); 294 abort(); /* bug */ 295} 296 297static void 298usage(void) 299{ 300 fprintf(stderr, "usage: unifdef [-cdeklnst] [-Ipath]" 301 " [-Dsym[=val]] [-Usym] [-iDsym[=val]] [-iUsym] ... [file]\n"); 302 exit(2); 303} 304 305/* 306 * A state transition function alters the global #if processing state 307 * in a particular way. The table below is indexed by the current 308 * processing state and the type of the current line. 309 * 310 * Nesting is handled by keeping a stack of states; some transition 311 * functions increase or decrease the depth. They also maintain the 312 * ignore state on a stack. In some complicated cases they have to 313 * alter the preprocessor directive, as follows. 314 * 315 * When we have processed a group that starts off with a known-false 316 * #if/#elif sequence (which has therefore been deleted) followed by a 317 * #elif that we don't understand and therefore must keep, we edit the 318 * latter into a #if to keep the nesting correct. 319 * 320 * When we find a true #elif in a group, the following block will 321 * always be kept and the rest of the sequence after the next #elif or 322 * #else will be discarded. We edit the #elif into a #else and the 323 * following directive to #endif since this has the desired behaviour. 324 * 325 * "Dodgy" directives are split across multiple lines, the most common 326 * example being a multi-line comment hanging off the right of the 327 * directive. We can handle them correctly only if there is no change 328 * from printing to dropping (or vice versa) caused by that directive. 329 * If the directive is the first of a group we have a choice between 330 * failing with an error, or passing it through unchanged instead of 331 * evaluating it. The latter is not the default to avoid questions from 332 * users about unifdef unexpectedly leaving behind preprocessor directives. 333 */ 334typedef void state_fn(void); 335 336/* report an error */ 337static void Eelif (void) { error("Inappropriate #elif"); } 338static void Eelse (void) { error("Inappropriate #else"); } 339static void Eendif(void) { error("Inappropriate #endif"); } 340static void Eeof (void) { error("Premature EOF"); } 341static void Eioccc(void) { error("Obfuscated preprocessor control line"); } 342/* plain line handling */ 343static void print (void) { flushline(true); } 344static void drop (void) { flushline(false); } 345/* output lacks group's start line */ 346static void Strue (void) { drop(); ignoreoff(); state(IS_TRUE_PREFIX); } 347static void Sfalse(void) { drop(); ignoreoff(); state(IS_FALSE_PREFIX); } 348static void Selse (void) { drop(); state(IS_TRUE_ELSE); } 349/* print/pass this block */ 350static void Pelif (void) { print(); ignoreoff(); state(IS_PASS_MIDDLE); } 351static void Pelse (void) { print(); state(IS_PASS_ELSE); } 352static void Pendif(void) { print(); unnest(); } 353/* discard this block */ 354static void Dfalse(void) { drop(); ignoreoff(); state(IS_FALSE_TRAILER); } 355static void Delif (void) { drop(); ignoreoff(); state(IS_FALSE_MIDDLE); } 356static void Delse (void) { drop(); state(IS_FALSE_ELSE); } 357static void Dendif(void) { drop(); unnest(); } 358/* first line of group */ 359static void Fdrop (void) { nest(); Dfalse(); } 360static void Fpass (void) { nest(); Pelif(); } 361static void Ftrue (void) { nest(); Strue(); } 362static void Ffalse(void) { nest(); Sfalse(); } 363/* variable pedantry for obfuscated lines */ 364static void Oiffy (void) { if (!iocccok) Eioccc(); Fpass(); ignoreon(); } 365static void Oif (void) { if (!iocccok) Eioccc(); Fpass(); } 366static void Oelif (void) { if (!iocccok) Eioccc(); Pelif(); } 367/* ignore comments in this block */ 368static void Idrop (void) { Fdrop(); ignoreon(); } 369static void Itrue (void) { Ftrue(); ignoreon(); } 370static void Ifalse(void) { Ffalse(); ignoreon(); } 371/* edit this line */ 372static void Mpass (void) { strncpy(keyword, "if ", 4); Pelif(); } 373static void Mtrue (void) { keywordedit("else\n"); state(IS_TRUE_MIDDLE); } 374static void Melif (void) { keywordedit("endif\n"); state(IS_FALSE_TRAILER); } 375static void Melse (void) { keywordedit("endif\n"); state(IS_FALSE_ELSE); } 376 377static state_fn * const trans_table[IS_COUNT][LT_COUNT] = { 378/* IS_OUTSIDE */ 379{ Itrue, Ifalse,Fpass, Ftrue, Ffalse,Eelif, Eelif, Eelif, Eelse, Eendif, 380 Oiffy, Oiffy, Fpass, Oif, Oif, Eelif, Eelif, Eelif, Eelse, Eendif, 381 print, done }, 382/* IS_FALSE_PREFIX */ 383{ Idrop, Idrop, Fdrop, Fdrop, Fdrop, Mpass, Strue, Sfalse,Selse, Dendif, 384 Idrop, Idrop, Fdrop, Fdrop, Fdrop, Mpass, Eioccc,Eioccc,Eioccc,Eioccc, 385 drop, Eeof }, 386/* IS_TRUE_PREFIX */ 387{ Itrue, Ifalse,Fpass, Ftrue, Ffalse,Dfalse,Dfalse,Dfalse,Delse, Dendif, 388 Oiffy, Oiffy, Fpass, Oif, Oif, Eioccc,Eioccc,Eioccc,Eioccc,Eioccc, 389 print, Eeof }, 390/* IS_PASS_MIDDLE */ 391{ Itrue, Ifalse,Fpass, Ftrue, Ffalse,Pelif, Mtrue, Delif, Pelse, Pendif, 392 Oiffy, Oiffy, Fpass, Oif, Oif, Pelif, Oelif, Oelif, Pelse, Pendif, 393 print, Eeof }, 394/* IS_FALSE_MIDDLE */ 395{ Idrop, Idrop, Fdrop, Fdrop, Fdrop, Pelif, Mtrue, Delif, Pelse, Pendif, 396 Idrop, Idrop, Fdrop, Fdrop, Fdrop, Eioccc,Eioccc,Eioccc,Eioccc,Eioccc, 397 drop, Eeof }, 398/* IS_TRUE_MIDDLE */ 399{ Itrue, Ifalse,Fpass, Ftrue, Ffalse,Melif, Melif, Melif, Melse, Pendif, 400 Oiffy, Oiffy, Fpass, Oif, Oif, Eioccc,Eioccc,Eioccc,Eioccc,Pendif, 401 print, Eeof }, 402/* IS_PASS_ELSE */ 403{ Itrue, Ifalse,Fpass, Ftrue, Ffalse,Eelif, Eelif, Eelif, Eelse, Pendif, 404 Oiffy, Oiffy, Fpass, Oif, Oif, Eelif, Eelif, Eelif, Eelse, Pendif, 405 print, Eeof }, 406/* IS_FALSE_ELSE */ 407{ Idrop, Idrop, Fdrop, Fdrop, Fdrop, Eelif, Eelif, Eelif, Eelse, Dendif, 408 Idrop, Idrop, Fdrop, Fdrop, Fdrop, Eelif, Eelif, Eelif, Eelse, Eioccc, 409 drop, Eeof }, 410/* IS_TRUE_ELSE */ 411{ Itrue, Ifalse,Fpass, Ftrue, Ffalse,Eelif, Eelif, Eelif, Eelse, Dendif, 412 Oiffy, Oiffy, Fpass, Oif, Oif, Eelif, Eelif, Eelif, Eelse, Eioccc, 413 print, Eeof }, 414/* IS_FALSE_TRAILER */ 415{ Idrop, Idrop, Fdrop, Fdrop, Fdrop, Dfalse,Dfalse,Dfalse,Delse, Dendif, 416 Idrop, Idrop, Fdrop, Fdrop, Fdrop, Dfalse,Dfalse,Dfalse,Delse, Eioccc, 417 drop, Eeof } 418/*TRUEI FALSEI IF TRUE FALSE ELIF ELTRUE ELFALSE ELSE ENDIF 419 TRUEI FALSEI IF TRUE FALSE ELIF ELTRUE ELFALSE ELSE ENDIF (DODGY) 420 PLAIN EOF */ 421}; 422 423/* 424 * State machine utility functions 425 */ 426static void 427done(void) 428{ 429 if (incomment) 430 error("EOF in comment"); 431 exit(exitstat); 432} 433static void 434ignoreoff(void) 435{ 436 if (depth == 0) 437 abort(); /* bug */ 438 ignoring[depth] = ignoring[depth-1]; 439} 440static void 441ignoreon(void) 442{ 443 ignoring[depth] = true; 444} 445static void 446keywordedit(const char *replacement) 447{ 448 size_t size = tline + sizeof(tline) - keyword; 449 char *dst = keyword; 450 const char *src = replacement; 451 if (size != 0) { 452 while ((--size != 0) && (*src != '\0')) 453 *dst++ = *src++; 454 *dst = '\0'; 455 } 456 print(); 457} 458static void 459nest(void) 460{ 461 depth += 1; 462 if (depth >= MAXDEPTH) 463 error("Too many levels of nesting"); 464 stifline[depth] = linenum; 465} 466static void 467unnest(void) 468{ 469 if (depth == 0) 470 abort(); /* bug */ 471 depth -= 1; 472} 473static void 474state(Ifstate is) 475{ 476 ifstate[depth] = is; 477} 478 479/* 480 * Write a line to the output or not, according to command line options. 481 */ 482static void 483flushline(bool keep) 484{ 485 if (symlist) 486 return; 487 if (keep ^ complement) { 488 if (lnnum && delcount > 0) 489 printf("#line %d\n", linenum); 490 fputs(tline, stdout); 491 delcount = 0; 492 } else { 493 if (lnblank) 494 putc('\n', stdout); 495 exitstat = 1; 496 delcount += 1; 497 } 498} 499 500/* 501 * The driver for the state machine. 502 */ 503static void 504process(void) 505{ 506 Linetype lineval; 507 508 for (;;) { 509 linenum++; 510 lineval = getline(); 511 trans_table[ifstate[depth]][lineval](); 512 debug("process %s -> %s depth %d", 513 linetype_name[lineval], 514 ifstate_name[ifstate[depth]], depth); 515 } 516} 517 518/* 519 * Parse a line and determine its type. We keep the preprocessor line 520 * parser state between calls in the global variable linestate, with 521 * help from skipcomment(). 522 */ 523static Linetype 524getline(void) 525{ 526 const char *cp; 527 int cursym; 528 int kwlen; 529 Linetype retval; 530 Comment_state wascomment; 531 532 if (fgets(tline, MAXLINE, input) == NULL) 533 return (LT_EOF); 534 retval = LT_PLAIN; 535 wascomment = incomment; 536 cp = skipcomment(tline); 537 if (linestate == LS_START) { 538 if (*cp == '#') { 539 linestate = LS_HASH; 540 cp = skipcomment(cp + 1); 541 } else if (*cp != '\0') 542 linestate = LS_DIRTY; 543 } 544 if (!incomment && linestate == LS_HASH) { 545 keyword = tline + (cp - tline); 546 cp = skipsym(cp); 547 kwlen = cp - keyword; 548 /* no way can we deal with a continuation inside a keyword */ 549 if (strncmp(cp, "\\\n", 2) == 0) 550 Eioccc(); 551 if (strlcmp("ifdef", keyword, kwlen) == 0 || 552 strlcmp("ifndef", keyword, kwlen) == 0) { 553 cp = skipcomment(cp); 554 if ((cursym = findsym(cp)) < 0) 555 retval = LT_IF; 556 else { 557 retval = (keyword[2] == 'n') 558 ? LT_FALSE : LT_TRUE; 559 if (value[cursym] == NULL) 560 retval = (retval == LT_TRUE) 561 ? LT_FALSE : LT_TRUE; 562 if (ignore[cursym]) 563 retval = (retval == LT_TRUE) 564 ? LT_TRUEI : LT_FALSEI; 565 } 566 cp = skipsym(cp); 567 } else if (strlcmp("if", keyword, kwlen) == 0) 568 retval = ifeval(&cp); 569 else if (strlcmp("elif", keyword, kwlen) == 0) 570 retval = ifeval(&cp) - LT_IF + LT_ELIF; 571 else if (strlcmp("else", keyword, kwlen) == 0) 572 retval = LT_ELSE; 573 else if (strlcmp("endif", keyword, kwlen) == 0) 574 retval = LT_ENDIF; 575 else { 576 linestate = LS_DIRTY; 577 retval = LT_PLAIN; 578 } 579 cp = skipcomment(cp); 580 if (*cp != '\0') { 581 linestate = LS_DIRTY; 582 if (retval == LT_TRUE || retval == LT_FALSE || 583 retval == LT_TRUEI || retval == LT_FALSEI) 584 retval = LT_IF; 585 if (retval == LT_ELTRUE || retval == LT_ELFALSE) 586 retval = LT_ELIF; 587 } 588 if (retval != LT_PLAIN && (wascomment || incomment)) { 589 retval += LT_DODGY; 590 if (incomment) 591 linestate = LS_DIRTY; 592 } 593 /* skipcomment should have changed the state */ 594 if (linestate == LS_HASH) 595 abort(); /* bug */ 596 } 597 if (linestate == LS_DIRTY) { 598 while (*cp != '\0') 599 cp = skipcomment(cp + 1); 600 } 601 debug("parser %s comment %s line", 602 comment_name[incomment], linestate_name[linestate]); 603 return (retval); 604} 605 606/* 607 * These are the binary operators that are supported by the expression 608 * evaluator. Note that if support for division is added then we also 609 * need short-circuiting booleans because of divide-by-zero. 610 */ 611static int op_lt(int a, int b) { return (a < b); } 612static int op_gt(int a, int b) { return (a > b); } 613static int op_le(int a, int b) { return (a <= b); } 614static int op_ge(int a, int b) { return (a >= b); } 615static int op_eq(int a, int b) { return (a == b); } 616static int op_ne(int a, int b) { return (a != b); } 617static int op_or(int a, int b) { return (a || b); } 618static int op_and(int a, int b) { return (a && b); } 619 620/* 621 * An evaluation function takes three arguments, as follows: (1) a pointer to 622 * an element of the precedence table which lists the operators at the current 623 * level of precedence; (2) a pointer to an integer which will receive the 624 * value of the expression; and (3) a pointer to a char* that points to the 625 * expression to be evaluated and that is updated to the end of the expression 626 * when evaluation is complete. The function returns LT_FALSE if the value of 627 * the expression is zero, LT_TRUE if it is non-zero, or LT_IF if the 628 * expression could not be evaluated. 629 */ 630struct ops; 631 632typedef Linetype eval_fn(const struct ops *, int *, const char **); 633 634static eval_fn eval_table, eval_unary; 635 636/* 637 * The precedence table. Expressions involving binary operators are evaluated 638 * in a table-driven way by eval_table. When it evaluates a subexpression it 639 * calls the inner function with its first argument pointing to the next 640 * element of the table. Innermost expressions have special non-table-driven 641 * handling. 642 */ 643static const struct ops { 644 eval_fn *inner; 645 struct op { 646 const char *str; 647 int (*fn)(int, int); 648 } op[5]; 649} eval_ops[] = { 650 { eval_table, { { "||", op_or } } }, 651 { eval_table, { { "&&", op_and } } }, 652 { eval_table, { { "==", op_eq }, 653 { "!=", op_ne } } }, 654 { eval_unary, { { "<=", op_le }, 655 { ">=", op_ge }, 656 { "<", op_lt }, 657 { ">", op_gt } } } 658}; 659 660/* 661 * Function for evaluating the innermost parts of expressions, 662 * viz. !expr (expr) defined(symbol) symbol number 663 * We reset the keepthis flag when we find a non-constant subexpression. 664 */ 665static Linetype 666eval_unary(const struct ops *ops, int *valp, const char **cpp) 667{ 668 const char *cp; 669 char *ep; 670 int sym; 671 672 cp = skipcomment(*cpp); 673 if (*cp == '!') { 674 debug("eval%d !", ops - eval_ops); 675 cp++; 676 if (eval_unary(ops, valp, &cp) == LT_IF) 677 return (LT_IF); 678 *valp = !*valp; 679 } else if (*cp == '(') { 680 cp++; 681 debug("eval%d (", ops - eval_ops); 682 if (eval_table(eval_ops, valp, &cp) == LT_IF) 683 return (LT_IF); 684 cp = skipcomment(cp); 685 if (*cp++ != ')') 686 return (LT_IF); 687 } else if (isdigit((unsigned char)*cp)) { 688 debug("eval%d number", ops - eval_ops); 689 *valp = strtol(cp, &ep, 0); 690 cp = skipsym(cp); 691 } else if (strncmp(cp, "defined", 7) == 0 && endsym(cp[7])) { 692 cp = skipcomment(cp+7); 693 debug("eval%d defined", ops - eval_ops); 694 if (*cp++ != '(') 695 return (LT_IF); 696 cp = skipcomment(cp); 697 sym = findsym(cp); 698 if (sym < 0) 699 return (LT_IF); 700 *valp = (value[sym] != NULL); 701 cp = skipsym(cp); 702 cp = skipcomment(cp); 703 if (*cp++ != ')') 704 return (LT_IF); 705 keepthis = false; 706 } else if (!endsym(*cp)) { 707 debug("eval%d symbol", ops - eval_ops); 708 sym = findsym(cp); 709 if (sym < 0) 710 return (LT_IF); 711 if (value[sym] == NULL) 712 *valp = 0; 713 else { 714 *valp = strtol(value[sym], &ep, 0); 715 if (*ep != '\0' || ep == value[sym]) 716 return (LT_IF); 717 } 718 cp = skipsym(cp); 719 keepthis = false; 720 } else { 721 debug("eval%d bad expr", ops - eval_ops); 722 return (LT_IF); 723 } 724 725 *cpp = cp; 726 debug("eval%d = %d", ops - eval_ops, *valp); 727 return (*valp ? LT_TRUE : LT_FALSE); 728} 729 730/* 731 * Table-driven evaluation of binary operators. 732 */ 733static Linetype 734eval_table(const struct ops *ops, int *valp, const char **cpp) 735{ 736 const struct op *op; 737 const char *cp; 738 int val; 739 740 debug("eval%d", ops - eval_ops); 741 cp = *cpp; 742 if (ops->inner(ops+1, valp, &cp) == LT_IF) 743 return (LT_IF); 744 for (;;) { 745 cp = skipcomment(cp); 746 for (op = ops->op; op->str != NULL; op++) 747 if (strncmp(cp, op->str, strlen(op->str)) == 0) 748 break; 749 if (op->str == NULL) 750 break; 751 cp += strlen(op->str); 752 debug("eval%d %s", ops - eval_ops, op->str); 753 if (ops->inner(ops+1, &val, &cp) == LT_IF) 754 return (LT_IF); 755 *valp = op->fn(*valp, val); 756 } 757 758 *cpp = cp; 759 debug("eval%d = %d", ops - eval_ops, *valp); 760 return (*valp ? LT_TRUE : LT_FALSE); 761} 762 763/* 764 * Evaluate the expression on a #if or #elif line. If we can work out 765 * the result we return LT_TRUE or LT_FALSE accordingly, otherwise we 766 * return just a generic LT_IF. 767 */ 768static Linetype 769ifeval(const char **cpp) 770{ 771 int ret; 772 int val; 773 774 debug("eval %s", *cpp); 775 keepthis = killconsts ? false : true; 776 ret = eval_table(eval_ops, &val, cpp); 777 debug("eval = %d", val); 778 return (keepthis ? LT_IF : ret); 779} 780 781static const char * 782skipcomment(const char *cp) 783{ 784 if (text || ignoring[depth]) { 785 for (; isspace((unsigned char)*cp); cp++) 786 if (*cp == '\n') 787 linestate = LS_START; 788 return (cp); 789 } 790 while (*cp != '\0') 791 /* don't reset to LS_START after a line continuation */ 792 if (strncmp(cp, "\\\n", 2) == 0) 793 cp += 2; 794 else switch (incomment) { 795 case NO_COMMENT: 796 if (strncmp(cp, "/\\\n", 3) == 0) { 797 incomment = STARTING_COMMENT; 798 cp += 3; 799 } else if (strncmp(cp, "/*", 2) == 0) { 800 incomment = C_COMMENT; 801 cp += 2; 802 } else if (strncmp(cp, "//", 2) == 0) { 803 incomment = CXX_COMMENT; 804 cp += 2; 805 } else if (strncmp(cp, "\'", 1) == 0) { 806 incomment = CHAR_LITERAL; 807 linestate = LS_DIRTY; 808 cp += 1; 809 } else if (strncmp(cp, "\"", 1) == 0) { 810 incomment = STRING_LITERAL; 811 linestate = LS_DIRTY; 812 cp += 1; 813 } else if (strncmp(cp, "\n", 1) == 0) { 814 linestate = LS_START; 815 cp += 1; 816 } else if (strchr(" \t", *cp) != NULL) { 817 cp += 1; 818 } else 819 return (cp); 820 continue; 821 case CXX_COMMENT: 822 if (strncmp(cp, "\n", 1) == 0) { 823 incomment = NO_COMMENT; 824 linestate = LS_START; 825 } 826 cp += 1; 827 continue; 828 case CHAR_LITERAL: 829 case STRING_LITERAL: 830 if ((incomment == CHAR_LITERAL && cp[0] == '\'') || 831 (incomment == STRING_LITERAL && cp[0] == '\"')) { 832 incomment = NO_COMMENT; 833 cp += 1; 834 } else if (cp[0] == '\\') { 835 if (cp[1] == '\0') 836 cp += 1; 837 else 838 cp += 2; 839 } else if (strncmp(cp, "\n", 1) == 0) { 840 if (incomment == CHAR_LITERAL) 841 error("unterminated char literal"); 842 else 843 error("unterminated string literal"); 844 } else 845 cp += 1; 846 continue; 847 case C_COMMENT: 848 if (strncmp(cp, "*\\\n", 3) == 0) { 849 incomment = FINISHING_COMMENT; 850 cp += 3; 851 } else if (strncmp(cp, "*/", 2) == 0) { 852 incomment = NO_COMMENT; 853 cp += 2; 854 } else 855 cp += 1; 856 continue; 857 case STARTING_COMMENT: 858 if (*cp == '*') { 859 incomment = C_COMMENT; 860 cp += 1; 861 } else if (*cp == '/') { 862 incomment = CXX_COMMENT; 863 cp += 1; 864 } else { 865 incomment = NO_COMMENT; 866 linestate = LS_DIRTY; 867 } 868 continue; 869 case FINISHING_COMMENT: 870 if (*cp == '/') { 871 incomment = NO_COMMENT; 872 cp += 1; 873 } else 874 incomment = C_COMMENT; 875 continue; 876 default: 877 abort(); /* bug */ 878 } 879 return (cp); 880} 881 882/* 883 * Skip over an identifier. 884 */ 885static const char * 886skipsym(const char *cp) 887{ 888 while (!endsym(*cp)) 889 ++cp; 890 return (cp); 891} 892 893/* 894 * Look for the symbol in the symbol table. If is is found, we return 895 * the symbol table index, else we return -1. 896 */ 897static int 898findsym(const char *str) 899{ 900 const char *cp; 901 int symind; 902 903 cp = skipsym(str); 904 if (cp == str) 905 return (-1); 906 if (symlist) { 907 printf("%.*s\n", (int)(cp-str), str); 908 /* we don't care about the value of the symbol */ 909 return (0); 910 } 911 for (symind = 0; symind < nsyms; ++symind) { 912 if (strlcmp(symname[symind], str, cp-str) == 0) { 913 debug("findsym %s %s", symname[symind], 914 value[symind] ? value[symind] : ""); 915 return (symind); 916 } 917 } 918 return (-1); 919} 920 921/* 922 * Add a symbol to the symbol table. 923 */ 924static void 925addsym(bool ignorethis, bool definethis, char *sym) 926{ 927 int symind; 928 char *val; 929 930 symind = findsym(sym); 931 if (symind < 0) { 932 if (nsyms >= MAXSYMS) 933 errx(2, "too many symbols"); 934 symind = nsyms++; 935 } 936 symname[symind] = sym; 937 ignore[symind] = ignorethis; 938 val = sym + (skipsym(sym) - sym); 939 if (definethis) { 940 if (*val == '=') { 941 value[symind] = val+1; 942 *val = '\0'; 943 } else if (*val == '\0') 944 value[symind] = ""; 945 else 946 usage(); 947 } else { 948 if (*val != '\0') 949 usage(); 950 value[symind] = NULL; 951 } 952} 953 954/* 955 * Compare s with n characters of t. 956 * The same as strncmp() except that it checks that s[n] == '\0'. 957 */ 958static int 959strlcmp(const char *s, const char *t, size_t n) 960{ 961 while (n-- && *t != '\0') 962 if (*s != *t) 963 return ((unsigned char)*s - (unsigned char)*t); 964 else 965 ++s, ++t; 966 return ((unsigned char)*s); 967} 968 969/* 970 * Diagnostics. 971 */ 972static void 973debug(const char *msg, ...) 974{ 975 va_list ap; 976 977 if (debugging) { 978 va_start(ap, msg); 979 vwarnx(msg, ap); 980 va_end(ap); 981 } 982} 983 984static void 985error(const char *msg) 986{ 987 if (depth == 0) 988 warnx("%s: %d: %s", filename, linenum, msg); 989 else 990 warnx("%s: %d: %s (#if line %d depth %d)", 991 filename, linenum, msg, stifline[depth], depth); 992 errx(2, "output may be truncated"); 993} 994