1/*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 2002 - 2015 Tony Finch <dot@dotat.at> 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 18 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 25 * SUCH DAMAGE. 26 */ 27 28/* 29 * unifdef - remove ifdef'ed lines 30 * 31 * This code was derived from software contributed to Berkeley by Dave Yost. 32 * It was rewritten to support ANSI C by Tony Finch. The original version 33 * of unifdef carried the 4-clause BSD copyright licence. None of its code 34 * remains in this version (though some of the names remain) so it now 35 * carries a more liberal licence. 36 * 37 * Wishlist: 38 * provide an option which will append the name of the 39 * appropriate symbol after #else's and #endif's 40 * provide an option which will check symbols after 41 * #else's and #endif's to see that they match their 42 * corresponding #ifdef or #ifndef 43 * 44 * These require better buffer handling, which would also make 45 * it possible to handle all "dodgy" directives correctly. 46 */ 47 48#include <sys/param.h> 49#include <sys/stat.h> 50#include <sys/tree.h> 51 52#include <assert.h> 53#include <ctype.h> 54#include <err.h> 55#include <stdarg.h> 56#include <stdbool.h> 57#include <stdio.h> 58#include <stdlib.h> 59#include <string.h> 60#include <unistd.h> 61 62static const char copyright[] = 63 "@(#) $Version: unifdef-2.11 $\n" 64 "@(#) $FreeBSD$\n" 65 "@(#) $Author: Tony Finch (dot@dotat.at) $\n" 66 "@(#) $URL: https://dotat.at/prog/unifdef $\n" 67; 68 69/* types of input lines: */ 70typedef enum { 71 LT_TRUEI, /* a true #if with ignore flag */ 72 LT_FALSEI, /* a false #if with ignore flag */ 73 LT_IF, /* an unknown #if */ 74 LT_TRUE, /* a true #if */ 75 LT_FALSE, /* a false #if */ 76 LT_ELIF, /* an unknown #elif */ 77 LT_ELTRUE, /* a true #elif */ 78 LT_ELFALSE, /* a false #elif */ 79 LT_ELSE, /* #else */ 80 LT_ENDIF, /* #endif */ 81 LT_DODGY, /* flag: directive is not on one line */ 82 LT_DODGY_LAST = LT_DODGY + LT_ENDIF, 83 LT_PLAIN, /* ordinary line */ 84 LT_EOF, /* end of file */ 85 LT_ERROR, /* unevaluable #if */ 86 LT_COUNT 87} Linetype; 88 89static char const * const linetype_name[] = { 90 "TRUEI", "FALSEI", "IF", "TRUE", "FALSE", 91 "ELIF", "ELTRUE", "ELFALSE", "ELSE", "ENDIF", 92 "DODGY TRUEI", "DODGY FALSEI", 93 "DODGY IF", "DODGY TRUE", "DODGY FALSE", 94 "DODGY ELIF", "DODGY ELTRUE", "DODGY ELFALSE", 95 "DODGY ELSE", "DODGY ENDIF", 96 "PLAIN", "EOF", "ERROR" 97}; 98 99#define linetype_if2elif(lt) ((Linetype)(lt - LT_IF + LT_ELIF)) 100#define linetype_2dodgy(lt) ((Linetype)(lt + LT_DODGY)) 101 102/* state of #if processing */ 103typedef enum { 104 IS_OUTSIDE, 105 IS_FALSE_PREFIX, /* false #if followed by false #elifs */ 106 IS_TRUE_PREFIX, /* first non-false #(el)if is true */ 107 IS_PASS_MIDDLE, /* first non-false #(el)if is unknown */ 108 IS_FALSE_MIDDLE, /* a false #elif after a pass state */ 109 IS_TRUE_MIDDLE, /* a true #elif after a pass state */ 110 IS_PASS_ELSE, /* an else after a pass state */ 111 IS_FALSE_ELSE, /* an else after a true state */ 112 IS_TRUE_ELSE, /* an else after only false states */ 113 IS_FALSE_TRAILER, /* #elifs after a true are false */ 114 IS_COUNT 115} Ifstate; 116 117static char const * const ifstate_name[] = { 118 "OUTSIDE", "FALSE_PREFIX", "TRUE_PREFIX", 119 "PASS_MIDDLE", "FALSE_MIDDLE", "TRUE_MIDDLE", 120 "PASS_ELSE", "FALSE_ELSE", "TRUE_ELSE", 121 "FALSE_TRAILER" 122}; 123 124/* state of comment parser */ 125typedef enum { 126 NO_COMMENT = false, /* outside a comment */ 127 C_COMMENT, /* in a comment like this one */ 128 CXX_COMMENT, /* between // and end of line */ 129 STARTING_COMMENT, /* just after slash-backslash-newline */ 130 FINISHING_COMMENT, /* star-backslash-newline in a C comment */ 131 CHAR_LITERAL, /* inside '' */ 132 STRING_LITERAL /* inside "" */ 133} Comment_state; 134 135static char const * const comment_name[] = { 136 "NO", "C", "CXX", "STARTING", "FINISHING", "CHAR", "STRING" 137}; 138 139/* state of preprocessor line parser */ 140typedef enum { 141 LS_START, /* only space and comments on this line */ 142 LS_HASH, /* only space, comments, and a hash */ 143 LS_DIRTY /* this line can't be a preprocessor line */ 144} Line_state; 145 146static char const * const linestate_name[] = { 147 "START", "HASH", "DIRTY" 148}; 149 150/* 151 * Minimum translation limits from ISO/IEC 9899:1999 5.2.4.1 152 */ 153#define MAXDEPTH 64 /* maximum #if nesting */ 154#define MAXLINE 4096 /* maximum length of line */ 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 * C17/18 allow 63 characters per macro name, but up to 127 arbitrarily large 164 * parameters. 165 */ 166struct macro { 167 RB_ENTRY(macro) entry; 168 const char *name; 169 const char *value; 170 bool ignore; /* -iDsym or -iUsym */ 171}; 172 173static int 174macro_cmp(struct macro *a, struct macro *b) 175{ 176 return (strcmp(a->name, b->name)); 177} 178 179static RB_HEAD(MACROMAP, macro) macro_tree = RB_INITIALIZER(¯o_tree); 180RB_GENERATE_STATIC(MACROMAP, macro, entry, macro_cmp); 181 182/* 183 * Globals. 184 */ 185 186static bool compblank; /* -B: compress blank lines */ 187static bool lnblank; /* -b: blank deleted lines */ 188static bool complement; /* -c: do the complement */ 189static bool debugging; /* -d: debugging reports */ 190static bool inplace; /* -m: modify in place */ 191static bool iocccok; /* -e: fewer IOCCC errors */ 192static bool strictlogic; /* -K: keep ambiguous #ifs */ 193static bool killconsts; /* -k: eval constant #ifs */ 194static bool lnnum; /* -n: add #line directives */ 195static bool symlist; /* -s: output symbol list */ 196static bool symdepth; /* -S: output symbol depth */ 197static bool text; /* -t: this is a text file */ 198 199static FILE *input; /* input file pointer */ 200static const char *filename; /* input file name */ 201static int linenum; /* current line number */ 202static const char *linefile; /* file name for #line */ 203static FILE *output; /* output file pointer */ 204static const char *ofilename; /* output file name */ 205static const char *backext; /* backup extension */ 206static char *tempname; /* avoid splatting input */ 207 208static char tline[MAXLINE+EDITSLOP];/* input buffer plus space */ 209static char *keyword; /* used for editing #elif's */ 210 211/* 212 * When processing a file, the output's newline style will match the 213 * input's, and unifdef correctly handles CRLF or LF endings whatever 214 * the platform's native style. The stdio streams are opened in binary 215 * mode to accommodate platforms whose native newline style is CRLF. 216 * When the output isn't a processed input file (when it is error / 217 * debug / diagnostic messages) then unifdef uses native line endings. 218 */ 219 220static const char *newline; /* input file format */ 221static const char newline_unix[] = "\n"; 222static const char newline_crlf[] = "\r\n"; 223 224static Comment_state incomment; /* comment parser state */ 225static Line_state linestate; /* #if line parser state */ 226static Ifstate ifstate[MAXDEPTH]; /* #if processor state */ 227static bool ignoring[MAXDEPTH]; /* ignore comments state */ 228static int stifline[MAXDEPTH]; /* start of current #if */ 229static int depth; /* current #if nesting */ 230static int delcount; /* count of deleted lines */ 231static unsigned blankcount; /* count of blank lines */ 232static unsigned blankmax; /* maximum recent blankcount */ 233static bool constexpr; /* constant #if expression */ 234static bool zerosyms; /* to format symdepth output */ 235static bool firstsym; /* ditto */ 236 237static int exitmode; /* exit status mode */ 238static int exitstat; /* program exit status */ 239static bool altered; /* was this file modified? */ 240 241static void addsym1(bool, bool, char *); 242static void addsym2(bool, const char *, const char *); 243static char *astrcat(const char *, const char *); 244static void cleantemp(void); 245static void closeio(void); 246static void debug(const char *, ...); 247static void debugsym(const char *, const struct macro *); 248static bool defundef(void); 249static void defundefile(const char *); 250static void done(void); 251static void error(const char *); 252static struct macro *findsym(const char **); 253static void flushline(bool); 254static void hashline(void); 255static void help(void); 256static Linetype ifeval(const char **); 257static void ignoreoff(void); 258static void ignoreon(void); 259static void indirectsym(void); 260static void keywordedit(const char *); 261static const char *matchsym(const char *, const char *); 262static void nest(void); 263static Linetype parseline(void); 264static void process(void); 265static void processinout(const char *, const char *); 266static const char *skipargs(const char *); 267static const char *skipcomment(const char *); 268static const char *skiphash(void); 269static const char *skipline(const char *); 270static const char *skipsym(const char *); 271static void state(Ifstate); 272static void unnest(void); 273static void usage(void); 274static void version(void); 275static const char *xstrdup(const char *, const char *); 276 277#define endsym(c) (!isalnum((unsigned char)c) && c != '_') 278 279static FILE * 280mktempmode(char *tmp, int mode) 281{ 282 int rc, fd; 283 284 mode &= (S_IRWXU|S_IRWXG|S_IRWXO); 285 fd = mkstemp(tmp); 286 if (fd < 0) 287 err(2, "can't create %s", tmp); 288 rc = fchmod(fd, mode); 289 if (rc < 0) 290 err(2, "can't fchmod %s mode=0o%o", tmp, mode); 291 return (fdopen(fd, "wb")); 292} 293 294/* 295 * The main program. 296 */ 297int 298main(int argc, char *argv[]) 299{ 300 int opt; 301 302 while ((opt = getopt(argc, argv, "i:D:U:f:I:M:o:x:bBcdehKklmnsStV")) != -1) 303 switch (opt) { 304 case 'i': /* treat stuff controlled by these symbols as text */ 305 /* 306 * For strict backwards-compatibility the U or D 307 * should be immediately after the -i but it doesn't 308 * matter much if we relax that requirement. 309 */ 310 opt = *optarg++; 311 if (opt == 'D') 312 addsym1(true, true, optarg); 313 else if (opt == 'U') 314 addsym1(true, false, optarg); 315 else 316 usage(); 317 break; 318 case 'D': /* define a symbol */ 319 addsym1(false, true, optarg); 320 break; 321 case 'U': /* undef a symbol */ 322 addsym1(false, false, optarg); 323 break; 324 case 'I': /* no-op for compatibility with cpp */ 325 break; 326 case 'b': /* blank deleted lines instead of omitting them */ 327 case 'l': /* backwards compatibility */ 328 lnblank = true; 329 break; 330 case 'B': /* compress blank lines around removed section */ 331 compblank = true; 332 break; 333 case 'c': /* treat -D as -U and vice versa */ 334 complement = true; 335 break; 336 case 'd': 337 debugging = true; 338 break; 339 case 'e': /* fewer errors from dodgy lines */ 340 iocccok = true; 341 break; 342 case 'f': /* definitions file */ 343 defundefile(optarg); 344 break; 345 case 'h': 346 help(); 347 break; 348 case 'K': /* keep ambiguous #ifs */ 349 strictlogic = true; 350 break; 351 case 'k': /* process constant #ifs */ 352 killconsts = true; 353 break; 354 case 'm': /* modify in place */ 355 inplace = true; 356 break; 357 case 'M': /* modify in place and keep backup */ 358 inplace = true; 359 if (strlen(optarg) > 0) 360 backext = optarg; 361 break; 362 case 'n': /* add #line directive after deleted lines */ 363 lnnum = true; 364 break; 365 case 'o': /* output to a file */ 366 ofilename = optarg; 367 break; 368 case 's': /* only output list of symbols that control #ifs */ 369 symlist = true; 370 break; 371 case 'S': /* list symbols with their nesting depth */ 372 symlist = symdepth = true; 373 break; 374 case 't': /* don't parse C comments */ 375 text = true; 376 break; 377 case 'V': 378 version(); 379 break; 380 case 'x': 381 exitmode = atoi(optarg); 382 if(exitmode < 0 || exitmode > 2) 383 usage(); 384 break; 385 default: 386 usage(); 387 } 388 argc -= optind; 389 argv += optind; 390 if (compblank && lnblank) 391 errx(2, "-B and -b are mutually exclusive"); 392 if (symlist && (ofilename != NULL || inplace || argc > 1)) 393 errx(2, "-s only works with one input file"); 394 if (argc > 1 && ofilename != NULL) 395 errx(2, "-o cannot be used with multiple input files"); 396 if (argc > 1 && !inplace) 397 errx(2, "multiple input files require -m or -M"); 398 if (argc == 0 && inplace) 399 errx(2, "-m requires an input file"); 400 if (argc == 0) 401 argc = 1; 402 if (argc == 1 && !inplace && ofilename == NULL) 403 ofilename = "-"; 404 indirectsym(); 405 406 atexit(cleantemp); 407 if (ofilename != NULL) 408 processinout(*argv, ofilename); 409 else while (argc-- > 0) { 410 processinout(*argv, *argv); 411 argv++; 412 } 413 switch(exitmode) { 414 case(0): exit(exitstat); 415 case(1): exit(!exitstat); 416 case(2): exit(0); 417 default: abort(); /* bug */ 418 } 419} 420 421/* 422 * File logistics. 423 */ 424static void 425processinout(const char *ifn, const char *ofn) 426{ 427 struct stat st; 428 429 if (ifn == NULL || strcmp(ifn, "-") == 0) { 430 filename = "[stdin]"; 431 linefile = NULL; 432 input = stdin; 433 } else { 434 filename = ifn; 435 linefile = ifn; 436 input = fopen(ifn, "rb"); 437 if (input == NULL) 438 err(2, "can't open %s", ifn); 439 } 440 if (strcmp(ofn, "-") == 0) { 441 output = stdout; 442 process(); 443 return; 444 } 445 if (stat(ofn, &st) < 0) { 446 output = fopen(ofn, "wb"); 447 if (output == NULL) 448 err(2, "can't create %s", ofn); 449 process(); 450 return; 451 } 452 453 tempname = astrcat(ofn, ".XXXXXX"); 454 output = mktempmode(tempname, st.st_mode); 455 if (output == NULL) 456 err(2, "can't create %s", tempname); 457 458 process(); 459 460 if (backext != NULL) { 461 char *backname = astrcat(ofn, backext); 462 if (rename(ofn, backname) < 0) 463 err(2, "can't rename \"%s\" to \"%s\"", ofn, backname); 464 free(backname); 465 } 466 /* leave file unmodified if unifdef made no changes */ 467 if (!altered && backext == NULL) { 468 if (remove(tempname) < 0) 469 warn("can't remove \"%s\"", tempname); 470 } else if (rename(tempname, ofn) < 0) 471 err(2, "can't rename \"%s\" to \"%s\"", tempname, ofn); 472 free(tempname); 473 tempname = NULL; 474} 475 476/* 477 * For cleaning up if there is an error. 478 */ 479static void 480cleantemp(void) 481{ 482 if (tempname != NULL) 483 remove(tempname); 484} 485 486/* 487 * Self-identification functions. 488 */ 489 490static void 491version(void) 492{ 493 const char *c = copyright; 494 for (;;) { 495 while (*++c != '$') 496 if (*c == '\0') 497 exit(0); 498 while (*++c != '$') 499 putc(*c, stderr); 500 putc('\n', stderr); 501 } 502} 503 504static void 505synopsis(FILE *fp) 506{ 507 fprintf(fp, 508 "usage: unifdef [-bBcdehKkmnsStV] [-x{012}] [-Mext] [-opath] \\\n" 509 " [-[i]Dsym[=val]] [-[i]Usym] [-fpath] ... [file] ...\n"); 510} 511 512static void 513usage(void) 514{ 515 synopsis(stderr); 516 exit(2); 517} 518 519static void 520help(void) 521{ 522 synopsis(stdout); 523 printf( 524 " -Dsym=val define preprocessor symbol with given value\n" 525 " -Dsym define preprocessor symbol with value 1\n" 526 " -Usym preprocessor symbol is undefined\n" 527 " -iDsym=val \\ ignore C strings and comments\n" 528 " -iDsym ) in sections controlled by these\n" 529 " -iUsym / preprocessor symbols\n" 530 " -fpath file containing #define and #undef directives\n" 531 " -b blank lines instead of deleting them\n" 532 " -B compress blank lines around deleted section\n" 533 " -c complement (invert) keep vs. delete\n" 534 " -d debugging mode\n" 535 " -e ignore multiline preprocessor directives\n" 536 " -h print help\n" 537 " -Ipath extra include file path (ignored)\n" 538 " -K disable && and || short-circuiting\n" 539 " -k process constant #if expressions\n" 540 " -Mext modify in place and keep backups\n" 541 " -m modify input files in place\n" 542 " -n add #line directives to output\n" 543 " -opath output file name\n" 544 " -S list #if control symbols with nesting\n" 545 " -s list #if control symbols\n" 546 " -t ignore C strings and comments\n" 547 " -V print version\n" 548 " -x{012} exit status mode\n" 549 ); 550 exit(0); 551} 552 553/* 554 * A state transition function alters the global #if processing state 555 * in a particular way. The table below is indexed by the current 556 * processing state and the type of the current line. 557 * 558 * Nesting is handled by keeping a stack of states; some transition 559 * functions increase or decrease the depth. They also maintain the 560 * ignore state on a stack. In some complicated cases they have to 561 * alter the preprocessor directive, as follows. 562 * 563 * When we have processed a group that starts off with a known-false 564 * #if/#elif sequence (which has therefore been deleted) followed by a 565 * #elif that we don't understand and therefore must keep, we edit the 566 * latter into a #if to keep the nesting correct. We use memcpy() to 567 * overwrite the 4 byte token "elif" with "if " without a '\0' byte. 568 * 569 * When we find a true #elif in a group, the following block will 570 * always be kept and the rest of the sequence after the next #elif or 571 * #else will be discarded. We edit the #elif into a #else and the 572 * following directive to #endif since this has the desired behaviour. 573 * 574 * "Dodgy" directives are split across multiple lines, the most common 575 * example being a multi-line comment hanging off the right of the 576 * directive. We can handle them correctly only if there is no change 577 * from printing to dropping (or vice versa) caused by that directive. 578 * If the directive is the first of a group we have a choice between 579 * failing with an error, or passing it through unchanged instead of 580 * evaluating it. The latter is not the default to avoid questions from 581 * users about unifdef unexpectedly leaving behind preprocessor directives. 582 */ 583typedef void state_fn(void); 584 585/* report an error */ 586static void Eelif (void) { error("Inappropriate #elif"); } 587static void Eelse (void) { error("Inappropriate #else"); } 588static void Eendif(void) { error("Inappropriate #endif"); } 589static void Eeof (void) { error("Premature EOF"); } 590static void Eioccc(void) { error("Obfuscated preprocessor control line"); } 591/* plain line handling */ 592static void print (void) { flushline(true); } 593static void drop (void) { flushline(false); } 594/* output lacks group's start line */ 595static void Strue (void) { drop(); ignoreoff(); state(IS_TRUE_PREFIX); } 596static void Sfalse(void) { drop(); ignoreoff(); state(IS_FALSE_PREFIX); } 597static void Selse (void) { drop(); state(IS_TRUE_ELSE); } 598/* print/pass this block */ 599static void Pelif (void) { print(); ignoreoff(); state(IS_PASS_MIDDLE); } 600static void Pelse (void) { print(); state(IS_PASS_ELSE); } 601static void Pendif(void) { print(); unnest(); } 602/* discard this block */ 603static void Dfalse(void) { drop(); ignoreoff(); state(IS_FALSE_TRAILER); } 604static void Delif (void) { drop(); ignoreoff(); state(IS_FALSE_MIDDLE); } 605static void Delse (void) { drop(); state(IS_FALSE_ELSE); } 606static void Dendif(void) { drop(); unnest(); } 607/* first line of group */ 608static void Fdrop (void) { nest(); Dfalse(); } 609static void Fpass (void) { nest(); Pelif(); } 610static void Ftrue (void) { nest(); Strue(); } 611static void Ffalse(void) { nest(); Sfalse(); } 612/* variable pedantry for obfuscated lines */ 613static void Oiffy (void) { if (!iocccok) Eioccc(); Fpass(); ignoreon(); } 614static void Oif (void) { if (!iocccok) Eioccc(); Fpass(); } 615static void Oelif (void) { if (!iocccok) Eioccc(); Pelif(); } 616/* ignore comments in this block */ 617static void Idrop (void) { Fdrop(); ignoreon(); } 618static void Itrue (void) { Ftrue(); ignoreon(); } 619static void Ifalse(void) { Ffalse(); ignoreon(); } 620/* modify this line */ 621static void Mpass (void) { memcpy(keyword, "if ", 4); Pelif(); } 622static void Mtrue (void) { keywordedit("else"); state(IS_TRUE_MIDDLE); } 623static void Melif (void) { keywordedit("endif"); state(IS_FALSE_TRAILER); } 624static void Melse (void) { keywordedit("endif"); state(IS_FALSE_ELSE); } 625 626static state_fn * const trans_table[IS_COUNT][LT_COUNT] = { 627/* IS_OUTSIDE */ 628{ Itrue, Ifalse,Fpass, Ftrue, Ffalse,Eelif, Eelif, Eelif, Eelse, Eendif, 629 Oiffy, Oiffy, Fpass, Oif, Oif, Eelif, Eelif, Eelif, Eelse, Eendif, 630 print, done, abort }, 631/* IS_FALSE_PREFIX */ 632{ Idrop, Idrop, Fdrop, Fdrop, Fdrop, Mpass, Strue, Sfalse,Selse, Dendif, 633 Idrop, Idrop, Fdrop, Fdrop, Fdrop, Mpass, Eioccc,Eioccc,Eioccc,Eioccc, 634 drop, Eeof, abort }, 635/* IS_TRUE_PREFIX */ 636{ Itrue, Ifalse,Fpass, Ftrue, Ffalse,Dfalse,Dfalse,Dfalse,Delse, Dendif, 637 Oiffy, Oiffy, Fpass, Oif, Oif, Eioccc,Eioccc,Eioccc,Eioccc,Eioccc, 638 print, Eeof, abort }, 639/* IS_PASS_MIDDLE */ 640{ Itrue, Ifalse,Fpass, Ftrue, Ffalse,Pelif, Mtrue, Delif, Pelse, Pendif, 641 Oiffy, Oiffy, Fpass, Oif, Oif, Pelif, Oelif, Oelif, Pelse, Pendif, 642 print, Eeof, abort }, 643/* IS_FALSE_MIDDLE */ 644{ Idrop, Idrop, Fdrop, Fdrop, Fdrop, Pelif, Mtrue, Delif, Pelse, Pendif, 645 Idrop, Idrop, Fdrop, Fdrop, Fdrop, Eioccc,Eioccc,Eioccc,Eioccc,Eioccc, 646 drop, Eeof, abort }, 647/* IS_TRUE_MIDDLE */ 648{ Itrue, Ifalse,Fpass, Ftrue, Ffalse,Melif, Melif, Melif, Melse, Pendif, 649 Oiffy, Oiffy, Fpass, Oif, Oif, Eioccc,Eioccc,Eioccc,Eioccc,Pendif, 650 print, Eeof, abort }, 651/* IS_PASS_ELSE */ 652{ Itrue, Ifalse,Fpass, Ftrue, Ffalse,Eelif, Eelif, Eelif, Eelse, Pendif, 653 Oiffy, Oiffy, Fpass, Oif, Oif, Eelif, Eelif, Eelif, Eelse, Pendif, 654 print, Eeof, abort }, 655/* IS_FALSE_ELSE */ 656{ Idrop, Idrop, Fdrop, Fdrop, Fdrop, Eelif, Eelif, Eelif, Eelse, Dendif, 657 Idrop, Idrop, Fdrop, Fdrop, Fdrop, Eelif, Eelif, Eelif, Eelse, Eioccc, 658 drop, Eeof, abort }, 659/* IS_TRUE_ELSE */ 660{ Itrue, Ifalse,Fpass, Ftrue, Ffalse,Eelif, Eelif, Eelif, Eelse, Dendif, 661 Oiffy, Oiffy, Fpass, Oif, Oif, Eelif, Eelif, Eelif, Eelse, Eioccc, 662 print, Eeof, abort }, 663/* IS_FALSE_TRAILER */ 664{ Idrop, Idrop, Fdrop, Fdrop, Fdrop, Dfalse,Dfalse,Dfalse,Delse, Dendif, 665 Idrop, Idrop, Fdrop, Fdrop, Fdrop, Dfalse,Dfalse,Dfalse,Delse, Eioccc, 666 drop, Eeof, abort } 667/*TRUEI FALSEI IF TRUE FALSE ELIF ELTRUE ELFALSE ELSE ENDIF 668 TRUEI FALSEI IF TRUE FALSE ELIF ELTRUE ELFALSE ELSE ENDIF (DODGY) 669 PLAIN EOF ERROR */ 670}; 671 672/* 673 * State machine utility functions 674 */ 675static void 676ignoreoff(void) 677{ 678 if (depth == 0) 679 abort(); /* bug */ 680 ignoring[depth] = ignoring[depth-1]; 681} 682static void 683ignoreon(void) 684{ 685 ignoring[depth] = true; 686} 687static void 688keywordedit(const char *replacement) 689{ 690 snprintf(keyword, tline + sizeof(tline) - keyword, 691 "%s%s", replacement, newline); 692 altered = true; 693 print(); 694} 695static void 696nest(void) 697{ 698 if (depth > MAXDEPTH-1) 699 abort(); /* bug */ 700 if (depth == MAXDEPTH-1) 701 error("Too many levels of nesting"); 702 depth += 1; 703 stifline[depth] = linenum; 704} 705static void 706unnest(void) 707{ 708 if (depth == 0) 709 abort(); /* bug */ 710 depth -= 1; 711} 712static void 713state(Ifstate is) 714{ 715 ifstate[depth] = is; 716} 717 718/* 719 * The last state transition function. When this is called, 720 * lineval == LT_EOF, so the process() loop will terminate. 721 */ 722static void 723done(void) 724{ 725 if (incomment) 726 error("EOF in comment"); 727 closeio(); 728} 729 730/* 731 * Write a line to the output or not, according to command line options. 732 * If writing fails, closeio() will print the error and exit. 733 */ 734static void 735flushline(bool keep) 736{ 737 if (symlist) 738 return; 739 if (keep ^ complement) { 740 bool blankline = tline[strspn(tline, " \t\r\n")] == '\0'; 741 if (blankline && compblank && blankcount != blankmax) { 742 delcount += 1; 743 blankcount += 1; 744 } else { 745 if (lnnum && delcount > 0) 746 hashline(); 747 if (fputs(tline, output) == EOF) 748 closeio(); 749 delcount = 0; 750 blankmax = blankcount = blankline ? blankcount + 1 : 0; 751 } 752 } else { 753 if (lnblank && fputs(newline, output) == EOF) 754 closeio(); 755 altered = true; 756 delcount += 1; 757 blankcount = 0; 758 } 759 if (debugging && fflush(output) == EOF) 760 closeio(); 761} 762 763/* 764 * Format of #line directives depends on whether we know the input filename. 765 */ 766static void 767hashline(void) 768{ 769 int e; 770 771 if (linefile == NULL) 772 e = fprintf(output, "#line %d%s", linenum, newline); 773 else 774 e = fprintf(output, "#line %d \"%s\"%s", 775 linenum, linefile, newline); 776 if (e < 0) 777 closeio(); 778} 779 780/* 781 * Flush the output and handle errors. 782 */ 783static void 784closeio(void) 785{ 786 /* Tidy up after findsym(). */ 787 if (symdepth && !zerosyms) 788 printf("\n"); 789 if (output != NULL && (ferror(output) || fclose(output) == EOF)) 790 err(2, "%s: can't write to output", filename); 791 fclose(input); 792} 793 794/* 795 * The driver for the state machine. 796 */ 797static void 798process(void) 799{ 800 Linetype lineval = LT_PLAIN; 801 /* When compressing blank lines, act as if the file 802 is preceded by a large number of blank lines. */ 803 blankmax = blankcount = 1000; 804 zerosyms = true; 805 newline = NULL; 806 linenum = 0; 807 altered = false; 808 while (lineval != LT_EOF) { 809 lineval = parseline(); 810 trans_table[ifstate[depth]][lineval](); 811 debug("process line %d %s -> %s depth %d", 812 linenum, linetype_name[lineval], 813 ifstate_name[ifstate[depth]], depth); 814 } 815 exitstat |= altered; 816} 817 818/* 819 * Parse a line and determine its type. We keep the preprocessor line 820 * parser state between calls in the global variable linestate, with 821 * help from skipcomment(). 822 */ 823static Linetype 824parseline(void) 825{ 826 const char *cp; 827 struct macro *cursym; 828 Linetype retval; 829 Comment_state wascomment; 830 831 wascomment = incomment; 832 cp = skiphash(); 833 if (cp == NULL) 834 return (LT_EOF); 835 if (newline == NULL) { 836 if (strrchr(tline, '\n') == strrchr(tline, '\r') + 1) 837 newline = newline_crlf; 838 else 839 newline = newline_unix; 840 } 841 if (*cp == '\0') { 842 retval = LT_PLAIN; 843 goto done; 844 } 845 keyword = tline + (cp - tline); 846 if ((cp = matchsym("ifdef", keyword)) != NULL || 847 (cp = matchsym("ifndef", keyword)) != NULL) { 848 cp = skipcomment(cp); 849 if ((cursym = findsym(&cp)) == NULL) 850 retval = LT_IF; 851 else { 852 retval = (keyword[2] == 'n') 853 ? LT_FALSE : LT_TRUE; 854 if (cursym->value == NULL) 855 retval = (retval == LT_TRUE) 856 ? LT_FALSE : LT_TRUE; 857 if (cursym->ignore) 858 retval = (retval == LT_TRUE) 859 ? LT_TRUEI : LT_FALSEI; 860 } 861 } else if ((cp = matchsym("if", keyword)) != NULL) 862 retval = ifeval(&cp); 863 else if ((cp = matchsym("elif", keyword)) != NULL) 864 retval = linetype_if2elif(ifeval(&cp)); 865 else if ((cp = matchsym("else", keyword)) != NULL) 866 retval = LT_ELSE; 867 else if ((cp = matchsym("endif", keyword)) != NULL) 868 retval = LT_ENDIF; 869 else { 870 cp = skipsym(keyword); 871 /* no way can we deal with a continuation inside a keyword */ 872 if (strncmp(cp, "\\\r\n", 3) == 0 || 873 strncmp(cp, "\\\n", 2) == 0) 874 Eioccc(); 875 cp = skipline(cp); 876 retval = LT_PLAIN; 877 goto done; 878 } 879 cp = skipcomment(cp); 880 if (*cp != '\0') { 881 cp = skipline(cp); 882 if (retval == LT_TRUE || retval == LT_FALSE || 883 retval == LT_TRUEI || retval == LT_FALSEI) 884 retval = LT_IF; 885 if (retval == LT_ELTRUE || retval == LT_ELFALSE) 886 retval = LT_ELIF; 887 } 888 /* the following can happen if the last line of the file lacks a 889 newline or if there is too much whitespace in a directive */ 890 if (linestate == LS_HASH) { 891 long len = cp - tline; 892 if (fgets(tline + len, MAXLINE - len, input) == NULL) { 893 if (ferror(input)) 894 err(2, "can't read %s", filename); 895 /* append the missing newline at eof */ 896 strcpy(tline + len, newline); 897 cp += strlen(newline); 898 linestate = LS_START; 899 } else { 900 linestate = LS_DIRTY; 901 } 902 } 903 if (retval != LT_PLAIN && (wascomment || linestate != LS_START)) { 904 retval = linetype_2dodgy(retval); 905 linestate = LS_DIRTY; 906 } 907done: 908 debug("parser line %d state %s comment %s line", linenum, 909 comment_name[incomment], linestate_name[linestate]); 910 return (retval); 911} 912 913/* 914 * These are the binary operators that are supported by the expression 915 * evaluator. 916 */ 917static Linetype op_strict(long *p, long v, Linetype at, Linetype bt) { 918 if(at == LT_IF || bt == LT_IF) return (LT_IF); 919 return (*p = v, v ? LT_TRUE : LT_FALSE); 920} 921static Linetype op_lt(long *p, Linetype at, long a, Linetype bt, long b) { 922 return op_strict(p, a < b, at, bt); 923} 924static Linetype op_gt(long *p, Linetype at, long a, Linetype bt, long b) { 925 return op_strict(p, a > b, at, bt); 926} 927static Linetype op_le(long *p, Linetype at, long a, Linetype bt, long b) { 928 return op_strict(p, a <= b, at, bt); 929} 930static Linetype op_ge(long *p, Linetype at, long a, Linetype bt, long b) { 931 return op_strict(p, a >= b, at, bt); 932} 933static Linetype op_eq(long *p, Linetype at, long a, Linetype bt, long b) { 934 return op_strict(p, a == b, at, bt); 935} 936static Linetype op_ne(long *p, Linetype at, long a, Linetype bt, long b) { 937 return op_strict(p, a != b, at, bt); 938} 939static Linetype op_or(long *p, Linetype at, long a, Linetype bt, long b) { 940 if (!strictlogic && (at == LT_TRUE || bt == LT_TRUE)) 941 return (*p = 1, LT_TRUE); 942 return op_strict(p, a || b, at, bt); 943} 944static Linetype op_and(long *p, Linetype at, long a, Linetype bt, long b) { 945 if (!strictlogic && (at == LT_FALSE || bt == LT_FALSE)) 946 return (*p = 0, LT_FALSE); 947 return op_strict(p, a && b, at, bt); 948} 949static Linetype op_blsh(long *p, Linetype at, long a, Linetype bt, long b) { 950 return op_strict(p, a << b, at, bt); 951} 952static Linetype op_brsh(long *p, Linetype at, long a, Linetype bt, long b) { 953 return op_strict(p, a >> b, at, bt); 954} 955static Linetype op_add(long *p, Linetype at, long a, Linetype bt, long b) { 956 return op_strict(p, a + b, at, bt); 957} 958static Linetype op_sub(long *p, Linetype at, long a, Linetype bt, long b) { 959 return op_strict(p, a - b, at, bt); 960} 961static Linetype op_mul(long *p, Linetype at, long a, Linetype bt, long b) { 962 return op_strict(p, a * b, at, bt); 963} 964static Linetype op_div(long *p, Linetype at, long a, Linetype bt, long b) { 965 if (bt != LT_TRUE) { 966 debug("eval division by zero"); 967 return (LT_ERROR); 968 } 969 return op_strict(p, a / b, at, bt); 970} 971static Linetype op_mod(long *p, Linetype at, long a, Linetype bt, long b) { 972 return op_strict(p, a % b, at, bt); 973} 974static Linetype op_bor(long *p, Linetype at, long a, Linetype bt, long b) { 975 return op_strict(p, a | b, at, bt); 976} 977static Linetype op_bxor(long *p, Linetype at, long a, Linetype bt, long b) { 978 return op_strict(p, a ^ b, at, bt); 979} 980static Linetype op_band(long *p, Linetype at, long a, Linetype bt, long b) { 981 return op_strict(p, a & b, at, bt); 982} 983 984/* 985 * An evaluation function takes three arguments, as follows: (1) a pointer to 986 * an element of the precedence table which lists the operators at the current 987 * level of precedence; (2) a pointer to an integer which will receive the 988 * value of the expression; and (3) a pointer to a char* that points to the 989 * expression to be evaluated and that is updated to the end of the expression 990 * when evaluation is complete. The function returns LT_FALSE if the value of 991 * the expression is zero, LT_TRUE if it is non-zero, LT_IF if the expression 992 * depends on an unknown symbol, or LT_ERROR if there is a parse failure. 993 */ 994struct ops; 995 996typedef Linetype eval_fn(const struct ops *, long *, const char **); 997 998static eval_fn eval_table, eval_unary; 999 1000/* 1001 * The precedence table. Expressions involving binary operators are evaluated 1002 * in a table-driven way by eval_table. When it evaluates a subexpression it 1003 * calls the inner function with its first argument pointing to the next 1004 * element of the table. Innermost expressions have special non-table-driven 1005 * handling. 1006 * 1007 * The stop characters help with lexical analysis: an operator is not 1008 * recognized if it is followed by one of the stop characters because 1009 * that would make it a different operator. 1010 */ 1011struct op { 1012 const char *str; 1013 Linetype (*fn)(long *, Linetype, long, Linetype, long); 1014 const char *stop; 1015}; 1016struct ops { 1017 eval_fn *inner; 1018 struct op op[5]; 1019}; 1020static const struct ops eval_ops[] = { 1021 { eval_table, { { "||", op_or, NULL } } }, 1022 { eval_table, { { "&&", op_and, NULL } } }, 1023 { eval_table, { { "|", op_bor, "|" } } }, 1024 { eval_table, { { "^", op_bxor, NULL } } }, 1025 { eval_table, { { "&", op_band, "&" } } }, 1026 { eval_table, { { "==", op_eq, NULL }, 1027 { "!=", op_ne, NULL } } }, 1028 { eval_table, { { "<=", op_le, NULL }, 1029 { ">=", op_ge, NULL }, 1030 { "<", op_lt, "<=" }, 1031 { ">", op_gt, ">=" } } }, 1032 { eval_table, { { "<<", op_blsh, NULL }, 1033 { ">>", op_brsh, NULL } } }, 1034 { eval_table, { { "+", op_add, NULL }, 1035 { "-", op_sub, NULL } } }, 1036 { eval_unary, { { "*", op_mul, NULL }, 1037 { "/", op_div, NULL }, 1038 { "%", op_mod, NULL } } }, 1039}; 1040 1041/* Current operator precedence level */ 1042static long prec(const struct ops *ops) 1043{ 1044 return (ops - eval_ops); 1045} 1046 1047/* 1048 * Function for evaluating the innermost parts of expressions, 1049 * viz. !expr (expr) number defined(symbol) symbol 1050 * We reset the constexpr flag in the last two cases. 1051 */ 1052static Linetype 1053eval_unary(const struct ops *ops, long *valp, const char **cpp) 1054{ 1055 const char *cp; 1056 char *ep; 1057 struct macro *sym; 1058 bool defparen; 1059 Linetype lt; 1060 1061 cp = skipcomment(*cpp); 1062 if (*cp == '!') { 1063 debug("eval%d !", prec(ops)); 1064 cp++; 1065 lt = eval_unary(ops, valp, &cp); 1066 if (lt == LT_ERROR) 1067 return (LT_ERROR); 1068 if (lt != LT_IF) { 1069 *valp = !*valp; 1070 lt = *valp ? LT_TRUE : LT_FALSE; 1071 } 1072 } else if (*cp == '~') { 1073 debug("eval%d ~", prec(ops)); 1074 cp++; 1075 lt = eval_unary(ops, valp, &cp); 1076 if (lt == LT_ERROR) 1077 return (LT_ERROR); 1078 if (lt != LT_IF) { 1079 *valp = ~(*valp); 1080 lt = *valp ? LT_TRUE : LT_FALSE; 1081 } 1082 } else if (*cp == '-') { 1083 debug("eval%d -", prec(ops)); 1084 cp++; 1085 lt = eval_unary(ops, valp, &cp); 1086 if (lt == LT_ERROR) 1087 return (LT_ERROR); 1088 if (lt != LT_IF) { 1089 *valp = -(*valp); 1090 lt = *valp ? LT_TRUE : LT_FALSE; 1091 } 1092 } else if (*cp == '(') { 1093 cp++; 1094 debug("eval%d (", prec(ops)); 1095 lt = eval_table(eval_ops, valp, &cp); 1096 if (lt == LT_ERROR) 1097 return (LT_ERROR); 1098 cp = skipcomment(cp); 1099 if (*cp++ != ')') 1100 return (LT_ERROR); 1101 } else if (isdigit((unsigned char)*cp)) { 1102 debug("eval%d number", prec(ops)); 1103 *valp = strtol(cp, &ep, 0); 1104 if (ep == cp) 1105 return (LT_ERROR); 1106 lt = *valp ? LT_TRUE : LT_FALSE; 1107 cp = ep; 1108 } else if (matchsym("defined", cp) != NULL) { 1109 cp = skipcomment(cp+7); 1110 if (*cp == '(') { 1111 cp = skipcomment(cp+1); 1112 defparen = true; 1113 } else { 1114 defparen = false; 1115 } 1116 sym = findsym(&cp); 1117 cp = skipcomment(cp); 1118 if (defparen && *cp++ != ')') { 1119 debug("eval%d defined missing ')'", prec(ops)); 1120 return (LT_ERROR); 1121 } 1122 if (sym == NULL) { 1123 debug("eval%d defined unknown", prec(ops)); 1124 lt = LT_IF; 1125 } else { 1126 debug("eval%d defined %s", prec(ops), sym->name); 1127 *valp = (sym->value != NULL); 1128 lt = *valp ? LT_TRUE : LT_FALSE; 1129 } 1130 constexpr = false; 1131 } else if (!endsym(*cp)) { 1132 debug("eval%d symbol", prec(ops)); 1133 sym = findsym(&cp); 1134 if (sym == NULL) { 1135 lt = LT_IF; 1136 cp = skipargs(cp); 1137 } else if (sym->value == NULL) { 1138 *valp = 0; 1139 lt = LT_FALSE; 1140 } else { 1141 *valp = strtol(sym->value, &ep, 0); 1142 if (*ep != '\0' || ep == sym->value) 1143 return (LT_ERROR); 1144 lt = *valp ? LT_TRUE : LT_FALSE; 1145 cp = skipargs(cp); 1146 } 1147 constexpr = false; 1148 } else { 1149 debug("eval%d bad expr", prec(ops)); 1150 return (LT_ERROR); 1151 } 1152 1153 *cpp = cp; 1154 debug("eval%d = %d", prec(ops), *valp); 1155 return (lt); 1156} 1157 1158/* 1159 * Table-driven evaluation of binary operators. 1160 */ 1161static Linetype 1162eval_table(const struct ops *ops, long *valp, const char **cpp) 1163{ 1164 const struct op *op; 1165 const char *cp; 1166 long val = 0; 1167 Linetype lt, rt; 1168 1169 debug("eval%d", prec(ops)); 1170 cp = *cpp; 1171 lt = ops->inner(ops+1, valp, &cp); 1172 if (lt == LT_ERROR) 1173 return (LT_ERROR); 1174 for (;;) { 1175 cp = skipcomment(cp); 1176 for (op = ops->op; op->str != NULL; op++) { 1177 if (strncmp(cp, op->str, strlen(op->str)) == 0) { 1178 /* assume only one-char operators have stop chars */ 1179 if (op->stop != NULL && cp[1] != '\0' && 1180 strchr(op->stop, cp[1]) != NULL) 1181 continue; 1182 else 1183 break; 1184 } 1185 } 1186 if (op->str == NULL) 1187 break; 1188 cp += strlen(op->str); 1189 debug("eval%d %s", prec(ops), op->str); 1190 rt = ops->inner(ops+1, &val, &cp); 1191 if (rt == LT_ERROR) 1192 return (LT_ERROR); 1193 lt = op->fn(valp, lt, *valp, rt, val); 1194 } 1195 1196 *cpp = cp; 1197 debug("eval%d = %d", prec(ops), *valp); 1198 debug("eval%d lt = %s", prec(ops), linetype_name[lt]); 1199 return (lt); 1200} 1201 1202/* 1203 * Evaluate the expression on a #if or #elif line. If we can work out 1204 * the result we return LT_TRUE or LT_FALSE accordingly, otherwise we 1205 * return just a generic LT_IF. 1206 */ 1207static Linetype 1208ifeval(const char **cpp) 1209{ 1210 Linetype ret; 1211 long val = 0; 1212 1213 debug("eval %s", *cpp); 1214 constexpr = killconsts ? false : true; 1215 ret = eval_table(eval_ops, &val, cpp); 1216 debug("eval = %d", val); 1217 return (constexpr ? LT_IF : ret == LT_ERROR ? LT_IF : ret); 1218} 1219 1220/* 1221 * Read a line and examine its initial part to determine if it is a 1222 * preprocessor directive. Returns NULL on EOF, or a pointer to a 1223 * preprocessor directive name, or a pointer to the zero byte at the 1224 * end of the line. 1225 */ 1226static const char * 1227skiphash(void) 1228{ 1229 const char *cp; 1230 1231 linenum++; 1232 if (fgets(tline, MAXLINE, input) == NULL) { 1233 if (ferror(input)) 1234 err(2, "can't read %s", filename); 1235 else 1236 return (NULL); 1237 } 1238 cp = skipcomment(tline); 1239 if (linestate == LS_START && *cp == '#') { 1240 linestate = LS_HASH; 1241 return (skipcomment(cp + 1)); 1242 } else if (*cp == '\0') { 1243 return (cp); 1244 } else { 1245 return (skipline(cp)); 1246 } 1247} 1248 1249/* 1250 * Mark a line dirty and consume the rest of it, keeping track of the 1251 * lexical state. 1252 */ 1253static const char * 1254skipline(const char *cp) 1255{ 1256 const char *pcp; 1257 if (*cp != '\0') 1258 linestate = LS_DIRTY; 1259 while (*cp != '\0') { 1260 cp = skipcomment(pcp = cp); 1261 if (pcp == cp) 1262 cp++; 1263 } 1264 return (cp); 1265} 1266 1267/* 1268 * Skip over comments, strings, and character literals and stop at the 1269 * next character position that is not whitespace. Between calls we keep 1270 * the comment state in the global variable incomment, and we also adjust 1271 * the global variable linestate when we see a newline. 1272 * XXX: doesn't cope with the buffer splitting inside a state transition. 1273 */ 1274static const char * 1275skipcomment(const char *cp) 1276{ 1277 if (text || ignoring[depth]) { 1278 for (; isspace((unsigned char)*cp); cp++) 1279 if (*cp == '\n') 1280 linestate = LS_START; 1281 return (cp); 1282 } 1283 while (*cp != '\0') 1284 /* don't reset to LS_START after a line continuation */ 1285 if (strncmp(cp, "\\\r\n", 3) == 0) 1286 cp += 3; 1287 else if (strncmp(cp, "\\\n", 2) == 0) 1288 cp += 2; 1289 else switch (incomment) { 1290 case NO_COMMENT: 1291 if (strncmp(cp, "/\\\r\n", 4) == 0) { 1292 incomment = STARTING_COMMENT; 1293 cp += 4; 1294 } else if (strncmp(cp, "/\\\n", 3) == 0) { 1295 incomment = STARTING_COMMENT; 1296 cp += 3; 1297 } else if (strncmp(cp, "/*", 2) == 0) { 1298 incomment = C_COMMENT; 1299 cp += 2; 1300 } else if (strncmp(cp, "//", 2) == 0) { 1301 incomment = CXX_COMMENT; 1302 cp += 2; 1303 } else if (strncmp(cp, "\'", 1) == 0) { 1304 incomment = CHAR_LITERAL; 1305 linestate = LS_DIRTY; 1306 cp += 1; 1307 } else if (strncmp(cp, "\"", 1) == 0) { 1308 incomment = STRING_LITERAL; 1309 linestate = LS_DIRTY; 1310 cp += 1; 1311 } else if (strncmp(cp, "\n", 1) == 0) { 1312 linestate = LS_START; 1313 cp += 1; 1314 } else if (strchr(" \r\t", *cp) != NULL) { 1315 cp += 1; 1316 } else 1317 return (cp); 1318 continue; 1319 case CXX_COMMENT: 1320 if (strncmp(cp, "\n", 1) == 0) { 1321 incomment = NO_COMMENT; 1322 linestate = LS_START; 1323 } 1324 cp += 1; 1325 continue; 1326 case CHAR_LITERAL: 1327 case STRING_LITERAL: 1328 if ((incomment == CHAR_LITERAL && cp[0] == '\'') || 1329 (incomment == STRING_LITERAL && cp[0] == '\"')) { 1330 incomment = NO_COMMENT; 1331 cp += 1; 1332 } else if (cp[0] == '\\') { 1333 if (cp[1] == '\0') 1334 cp += 1; 1335 else 1336 cp += 2; 1337 } else if (strncmp(cp, "\n", 1) == 0) { 1338 if (incomment == CHAR_LITERAL) 1339 error("Unterminated char literal"); 1340 else 1341 error("Unterminated string literal"); 1342 } else 1343 cp += 1; 1344 continue; 1345 case C_COMMENT: 1346 if (strncmp(cp, "*\\\r\n", 4) == 0) { 1347 incomment = FINISHING_COMMENT; 1348 cp += 4; 1349 } else if (strncmp(cp, "*\\\n", 3) == 0) { 1350 incomment = FINISHING_COMMENT; 1351 cp += 3; 1352 } else if (strncmp(cp, "*/", 2) == 0) { 1353 incomment = NO_COMMENT; 1354 cp += 2; 1355 } else 1356 cp += 1; 1357 continue; 1358 case STARTING_COMMENT: 1359 if (*cp == '*') { 1360 incomment = C_COMMENT; 1361 cp += 1; 1362 } else if (*cp == '/') { 1363 incomment = CXX_COMMENT; 1364 cp += 1; 1365 } else { 1366 incomment = NO_COMMENT; 1367 linestate = LS_DIRTY; 1368 } 1369 continue; 1370 case FINISHING_COMMENT: 1371 if (*cp == '/') { 1372 incomment = NO_COMMENT; 1373 cp += 1; 1374 } else 1375 incomment = C_COMMENT; 1376 continue; 1377 default: 1378 abort(); /* bug */ 1379 } 1380 return (cp); 1381} 1382 1383/* 1384 * Skip macro arguments. 1385 */ 1386static const char * 1387skipargs(const char *cp) 1388{ 1389 const char *ocp = cp; 1390 int level = 0; 1391 cp = skipcomment(cp); 1392 if (*cp != '(') 1393 return (cp); 1394 do { 1395 if (*cp == '(') 1396 level++; 1397 if (*cp == ')') 1398 level--; 1399 cp = skipcomment(cp+1); 1400 } while (level != 0 && *cp != '\0'); 1401 if (level == 0) 1402 return (cp); 1403 else 1404 /* Rewind and re-detect the syntax error later. */ 1405 return (ocp); 1406} 1407 1408/* 1409 * Skip over an identifier. 1410 */ 1411static const char * 1412skipsym(const char *cp) 1413{ 1414 while (!endsym(*cp)) 1415 ++cp; 1416 return (cp); 1417} 1418 1419/* 1420 * Skip whitespace and take a copy of any following identifier. 1421 */ 1422static const char * 1423getsym(const char **cpp) 1424{ 1425 const char *cp = *cpp, *sym; 1426 1427 cp = skipcomment(cp); 1428 cp = skipsym(sym = cp); 1429 if (cp == sym) 1430 return NULL; 1431 *cpp = cp; 1432 return (xstrdup(sym, cp)); 1433} 1434 1435/* 1436 * Check that s (a symbol) matches the start of t, and that the 1437 * following character in t is not a symbol character. Returns a 1438 * pointer to the following character in t if there is a match, 1439 * otherwise NULL. 1440 */ 1441static const char * 1442matchsym(const char *s, const char *t) 1443{ 1444 while (*s != '\0' && *t != '\0') 1445 if (*s != *t) 1446 return (NULL); 1447 else 1448 ++s, ++t; 1449 if (*s == '\0' && endsym(*t)) 1450 return(t); 1451 else 1452 return(NULL); 1453} 1454 1455/* 1456 * Look for the symbol in the symbol table. If it is found, we return 1457 * the symbol table index, else we return -1. 1458 */ 1459static struct macro * 1460findsym(const char **strp) 1461{ 1462 const char *str; 1463 char *strkey; 1464 struct macro key, *res; 1465 1466 str = *strp; 1467 *strp = skipsym(str); 1468 if (symlist) { 1469 if (*strp == str) 1470 return (NULL); 1471 if (symdepth && firstsym) 1472 printf("%s%3d", zerosyms ? "" : "\n", depth); 1473 firstsym = zerosyms = false; 1474 printf("%s%.*s%s", 1475 symdepth ? " " : "", 1476 (int)(*strp-str), str, 1477 symdepth ? "" : "\n"); 1478 /* we don't care about the value of the symbol */ 1479 return (NULL); 1480 } 1481 1482 /* 1483 * 'str' just points into the current mid-parse input and is not 1484 * nul-terminated. We know the length of the symbol, *strp - str, but 1485 * need to provide a nul-terminated lookup key for RB_FIND's comparison 1486 * function. Create one here. 1487 */ 1488 strkey = malloc(*strp - str + 1); 1489 memcpy(strkey, str, *strp - str); 1490 strkey[*strp - str] = 0; 1491 1492 key.name = strkey; 1493 res = RB_FIND(MACROMAP, ¯o_tree, &key); 1494 if (res != NULL) 1495 debugsym("findsym", res); 1496 1497 free(strkey); 1498 return (res); 1499} 1500 1501/* 1502 * Resolve indirect symbol values to their final definitions. 1503 */ 1504static void 1505indirectsym(void) 1506{ 1507 const char *cp; 1508 int changed; 1509 struct macro *sym, *ind; 1510 1511 do { 1512 changed = 0; 1513 RB_FOREACH(sym, MACROMAP, ¯o_tree) { 1514 if (sym->value == NULL) 1515 continue; 1516 cp = sym->value; 1517 ind = findsym(&cp); 1518 if (ind == NULL || ind == sym || 1519 *cp != '\0' || 1520 ind->value == NULL || 1521 ind->value == sym->value) 1522 continue; 1523 debugsym("indir...", sym); 1524 sym->value = ind->value; 1525 debugsym("...ectsym", sym); 1526 changed++; 1527 } 1528 } while (changed); 1529} 1530 1531/* 1532 * Add a symbol to the symbol table, specified with the format sym=val 1533 */ 1534static void 1535addsym1(bool ignorethis, bool definethis, char *symval) 1536{ 1537 const char *sym, *val; 1538 1539 sym = symval; 1540 val = skipsym(sym); 1541 if (definethis && *val == '=') { 1542 symval[val - sym] = '\0'; 1543 val = val + 1; 1544 } else if (*val == '\0') { 1545 val = definethis ? "1" : NULL; 1546 } else { 1547 usage(); 1548 } 1549 addsym2(ignorethis, sym, val); 1550} 1551 1552/* 1553 * Add a symbol to the symbol table. 1554 */ 1555static void 1556addsym2(bool ignorethis, const char *symname, const char *val) 1557{ 1558 const char *cp = symname; 1559 struct macro *sym, *r; 1560 1561 sym = findsym(&cp); 1562 if (sym == NULL) { 1563 sym = calloc(1, sizeof(*sym)); 1564 sym->ignore = ignorethis; 1565 sym->name = symname; 1566 sym->value = val; 1567 r = RB_INSERT(MACROMAP, ¯o_tree, sym); 1568 assert(r == NULL); 1569 } 1570 debugsym("addsym", sym); 1571} 1572 1573static void 1574debugsym(const char *why, const struct macro *sym) 1575{ 1576 debug("%s %s%c%s", why, sym->name, 1577 sym->value ? '=' : ' ', 1578 sym->value ? sym->value : "undef"); 1579} 1580 1581/* 1582 * Add symbols to the symbol table from a file containing 1583 * #define and #undef preprocessor directives. 1584 */ 1585static void 1586defundefile(const char *fn) 1587{ 1588 filename = fn; 1589 input = fopen(fn, "rb"); 1590 if (input == NULL) 1591 err(2, "can't open %s", fn); 1592 linenum = 0; 1593 while (defundef()) 1594 ; 1595 if (ferror(input)) 1596 err(2, "can't read %s", filename); 1597 else 1598 fclose(input); 1599 if (incomment) 1600 error("EOF in comment"); 1601} 1602 1603/* 1604 * Read and process one #define or #undef directive 1605 */ 1606static bool 1607defundef(void) 1608{ 1609 const char *cp, *kw, *sym, *val, *end; 1610 1611 cp = skiphash(); 1612 if (cp == NULL) 1613 return (false); 1614 if (*cp == '\0') 1615 goto done; 1616 /* strip trailing whitespace, and do a fairly rough check to 1617 avoid unsupported multi-line preprocessor directives */ 1618 end = cp + strlen(cp); 1619 while (end > tline && strchr(" \t\n\r", end[-1]) != NULL) 1620 --end; 1621 if (end > tline && end[-1] == '\\') 1622 Eioccc(); 1623 1624 kw = cp; 1625 if ((cp = matchsym("define", kw)) != NULL) { 1626 sym = getsym(&cp); 1627 if (sym == NULL) 1628 error("Missing macro name in #define"); 1629 if (*cp == '(') { 1630 val = "1"; 1631 } else { 1632 cp = skipcomment(cp); 1633 val = (cp < end) ? xstrdup(cp, end) : ""; 1634 } 1635 debug("#define"); 1636 addsym2(false, sym, val); 1637 } else if ((cp = matchsym("undef", kw)) != NULL) { 1638 sym = getsym(&cp); 1639 if (sym == NULL) 1640 error("Missing macro name in #undef"); 1641 cp = skipcomment(cp); 1642 debug("#undef"); 1643 addsym2(false, sym, NULL); 1644 } else { 1645 error("Unrecognized preprocessor directive"); 1646 } 1647 skipline(cp); 1648done: 1649 debug("parser line %d state %s comment %s line", linenum, 1650 comment_name[incomment], linestate_name[linestate]); 1651 return (true); 1652} 1653 1654/* 1655 * Concatenate two strings into new memory, checking for failure. 1656 */ 1657static char * 1658astrcat(const char *s1, const char *s2) 1659{ 1660 char *s; 1661 int len; 1662 size_t size; 1663 1664 len = snprintf(NULL, 0, "%s%s", s1, s2); 1665 if (len < 0) 1666 err(2, "snprintf"); 1667 size = (size_t)len + 1; 1668 s = (char *)malloc(size); 1669 if (s == NULL) 1670 err(2, "malloc"); 1671 snprintf(s, size, "%s%s", s1, s2); 1672 return (s); 1673} 1674 1675/* 1676 * Duplicate a segment of a string, checking for failure. 1677 */ 1678static const char * 1679xstrdup(const char *start, const char *end) 1680{ 1681 size_t n; 1682 char *s; 1683 1684 if (end < start) abort(); /* bug */ 1685 n = (size_t)(end - start) + 1; 1686 s = malloc(n); 1687 if (s == NULL) 1688 err(2, "malloc"); 1689 snprintf(s, n, "%s", start); 1690 return (s); 1691} 1692 1693/* 1694 * Diagnostics. 1695 */ 1696static void 1697debug(const char *msg, ...) 1698{ 1699 va_list ap; 1700 1701 if (debugging) { 1702 va_start(ap, msg); 1703 vwarnx(msg, ap); 1704 va_end(ap); 1705 } 1706} 1707 1708static void 1709error(const char *msg) 1710{ 1711 if (depth == 0) 1712 warnx("%s: %d: %s", filename, linenum, msg); 1713 else 1714 warnx("%s: %d: %s (#if line %d depth %d)", 1715 filename, linenum, msg, stifline[depth], depth); 1716 closeio(); 1717 errx(2, "Output may be truncated"); 1718} 1719