1 /*-
2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3 *
4 * Copyright (c) 2002 - 2015 Tony Finch <[email protected]>
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 "unifdef.h"
49
50 static const char copyright[] =
51 "@(#) $Version: unifdef-2.11 $\n"
52 "@(#) $FreeBSD$\n"
53 "@(#) $Author: Tony Finch ([email protected]) $\n"
54 "@(#) $URL: https://dotat.at/prog/unifdef $\n"
55 ;
56
57 /* types of input lines: */
58 typedef enum {
59 LT_TRUEI, /* a true #if with ignore flag */
60 LT_FALSEI, /* a false #if with ignore flag */
61 LT_IF, /* an unknown #if */
62 LT_TRUE, /* a true #if */
63 LT_FALSE, /* a false #if */
64 LT_ELIF, /* an unknown #elif */
65 LT_ELTRUE, /* a true #elif */
66 LT_ELFALSE, /* a false #elif */
67 LT_ELSE, /* #else */
68 LT_ENDIF, /* #endif */
69 LT_DODGY, /* flag: directive is not on one line */
70 LT_DODGY_LAST = LT_DODGY + LT_ENDIF,
71 LT_PLAIN, /* ordinary line */
72 LT_EOF, /* end of file */
73 LT_ERROR, /* unevaluable #if */
74 LT_COUNT
75 } Linetype;
76
77 static char const * const linetype_name[] = {
78 "TRUEI", "FALSEI", "IF", "TRUE", "FALSE",
79 "ELIF", "ELTRUE", "ELFALSE", "ELSE", "ENDIF",
80 "DODGY TRUEI", "DODGY FALSEI",
81 "DODGY IF", "DODGY TRUE", "DODGY FALSE",
82 "DODGY ELIF", "DODGY ELTRUE", "DODGY ELFALSE",
83 "DODGY ELSE", "DODGY ENDIF",
84 "PLAIN", "EOF", "ERROR"
85 };
86
87 #define linetype_if2elif(lt) ((Linetype)(lt - LT_IF + LT_ELIF))
88 #define linetype_2dodgy(lt) ((Linetype)(lt + LT_DODGY))
89
90 /* state of #if processing */
91 typedef enum {
92 IS_OUTSIDE,
93 IS_FALSE_PREFIX, /* false #if followed by false #elifs */
94 IS_TRUE_PREFIX, /* first non-false #(el)if is true */
95 IS_PASS_MIDDLE, /* first non-false #(el)if is unknown */
96 IS_FALSE_MIDDLE, /* a false #elif after a pass state */
97 IS_TRUE_MIDDLE, /* a true #elif after a pass state */
98 IS_PASS_ELSE, /* an else after a pass state */
99 IS_FALSE_ELSE, /* an else after a true state */
100 IS_TRUE_ELSE, /* an else after only false states */
101 IS_FALSE_TRAILER, /* #elifs after a true are false */
102 IS_COUNT
103 } Ifstate;
104
105 static char const * const ifstate_name[] = {
106 "OUTSIDE", "FALSE_PREFIX", "TRUE_PREFIX",
107 "PASS_MIDDLE", "FALSE_MIDDLE", "TRUE_MIDDLE",
108 "PASS_ELSE", "FALSE_ELSE", "TRUE_ELSE",
109 "FALSE_TRAILER"
110 };
111
112 /* state of comment parser */
113 typedef enum {
114 NO_COMMENT = false, /* outside a comment */
115 C_COMMENT, /* in a comment like this one */
116 CXX_COMMENT, /* between // and end of line */
117 STARTING_COMMENT, /* just after slash-backslash-newline */
118 FINISHING_COMMENT, /* star-backslash-newline in a C comment */
119 CHAR_LITERAL, /* inside '' */
120 STRING_LITERAL /* inside "" */
121 } Comment_state;
122
123 static char const * const comment_name[] = {
124 "NO", "C", "CXX", "STARTING", "FINISHING", "CHAR", "STRING"
125 };
126
127 /* state of preprocessor line parser */
128 typedef enum {
129 LS_START, /* only space and comments on this line */
130 LS_HASH, /* only space, comments, and a hash */
131 LS_DIRTY /* this line can't be a preprocessor line */
132 } Line_state;
133
134 static char const * const linestate_name[] = {
135 "START", "HASH", "DIRTY"
136 };
137
138 /*
139 * Minimum translation limits from ISO/IEC 9899:1999 5.2.4.1
140 */
141 #define MAXDEPTH 64 /* maximum #if nesting */
142 #define MAXLINE 4096 /* maximum length of line */
143 #define MAXSYMS 16384 /* maximum number of symbols */
144
145 /*
146 * Sometimes when editing a keyword the replacement text is longer, so
147 * we leave some space at the end of the tline buffer to accommodate this.
148 */
149 #define EDITSLOP 10
150
151 /*
152 * Globals.
153 */
154
155 static bool compblank; /* -B: compress blank lines */
156 static bool lnblank; /* -b: blank deleted lines */
157 static bool complement; /* -c: do the complement */
158 static bool debugging; /* -d: debugging reports */
159 static bool inplace; /* -m: modify in place */
160 static bool iocccok; /* -e: fewer IOCCC errors */
161 static bool strictlogic; /* -K: keep ambiguous #ifs */
162 static bool killconsts; /* -k: eval constant #ifs */
163 static bool lnnum; /* -n: add #line directives */
164 static bool symlist; /* -s: output symbol list */
165 static bool symdepth; /* -S: output symbol depth */
166 static bool text; /* -t: this is a text file */
167
168 static const char *symname[MAXSYMS]; /* symbol name */
169 static const char *value[MAXSYMS]; /* -Dsym=value */
170 static bool ignore[MAXSYMS]; /* -iDsym or -iUsym */
171 static int nsyms; /* number of symbols */
172
173 static FILE *input; /* input file pointer */
174 static const char *filename; /* input file name */
175 static int linenum; /* current line number */
176 static const char *linefile; /* file name for #line */
177 static FILE *output; /* output file pointer */
178 static const char *ofilename; /* output file name */
179 static const char *backext; /* backup extension */
180 static char *tempname; /* avoid splatting input */
181
182 static char tline[MAXLINE+EDITSLOP];/* input buffer plus space */
183 static char *keyword; /* used for editing #elif's */
184
185 /*
186 * When processing a file, the output's newline style will match the
187 * input's, and unifdef correctly handles CRLF or LF endings whatever
188 * the platform's native style. The stdio streams are opened in binary
189 * mode to accommodate platforms whose native newline style is CRLF.
190 * When the output isn't a processed input file (when it is error /
191 * debug / diagnostic messages) then unifdef uses native line endings.
192 */
193
194 static const char *newline; /* input file format */
195 static const char newline_unix[] = "\n";
196 static const char newline_crlf[] = "\r\n";
197
198 static Comment_state incomment; /* comment parser state */
199 static Line_state linestate; /* #if line parser state */
200 static Ifstate ifstate[MAXDEPTH]; /* #if processor state */
201 static bool ignoring[MAXDEPTH]; /* ignore comments state */
202 static int stifline[MAXDEPTH]; /* start of current #if */
203 static int depth; /* current #if nesting */
204 static int delcount; /* count of deleted lines */
205 static unsigned blankcount; /* count of blank lines */
206 static unsigned blankmax; /* maximum recent blankcount */
207 static bool constexpr; /* constant #if expression */
208 static bool zerosyms; /* to format symdepth output */
209 static bool firstsym; /* ditto */
210
211 static int exitmode; /* exit status mode */
212 static int exitstat; /* program exit status */
213 static bool altered; /* was this file modified? */
214
215 static void addsym1(bool, bool, char *);
216 static void addsym2(bool, const char *, const char *);
217 static char *astrcat(const char *, const char *);
218 static void cleantemp(void);
219 static void closeio(void);
220 static void debug(const char *, ...);
221 static void debugsym(const char *, int);
222 static bool defundef(void);
223 static void defundefile(const char *);
224 static void done(void);
225 static void error(const char *);
226 static int findsym(const char **);
227 static void flushline(bool);
228 static void hashline(void);
229 static void help(void);
230 static Linetype ifeval(const char **);
231 static void ignoreoff(void);
232 static void ignoreon(void);
233 static void indirectsym(void);
234 static void keywordedit(const char *);
235 static const char *matchsym(const char *, const char *);
236 static void nest(void);
237 static Linetype parseline(void);
238 static void process(void);
239 static void processinout(const char *, const char *);
240 static const char *skipargs(const char *);
241 static const char *skipcomment(const char *);
242 static const char *skiphash(void);
243 static const char *skipline(const char *);
244 static const char *skipsym(const char *);
245 static void state(Ifstate);
246 static void unnest(void);
247 static void usage(void);
248 static void version(void);
249 static const char *xstrdup(const char *, const char *);
250
251 #define endsym(c) (!isalnum((unsigned char)c) && c != '_')
252
253 /*
254 * The main program.
255 */
256 int
main(int argc,char * argv[])257 main(int argc, char *argv[])
258 {
259 int opt;
260
261 while ((opt = getopt(argc, argv, "i:D:U:f:I:M:o:x:bBcdehKklmnsStV")) != -1)
262 switch (opt) {
263 case 'i': /* treat stuff controlled by these symbols as text */
264 /*
265 * For strict backwards-compatibility the U or D
266 * should be immediately after the -i but it doesn't
267 * matter much if we relax that requirement.
268 */
269 opt = *optarg++;
270 if (opt == 'D')
271 addsym1(true, true, optarg);
272 else if (opt == 'U')
273 addsym1(true, false, optarg);
274 else
275 usage();
276 break;
277 case 'D': /* define a symbol */
278 addsym1(false, true, optarg);
279 break;
280 case 'U': /* undef a symbol */
281 addsym1(false, false, optarg);
282 break;
283 case 'I': /* no-op for compatibility with cpp */
284 break;
285 case 'b': /* blank deleted lines instead of omitting them */
286 case 'l': /* backwards compatibility */
287 lnblank = true;
288 break;
289 case 'B': /* compress blank lines around removed section */
290 compblank = true;
291 break;
292 case 'c': /* treat -D as -U and vice versa */
293 complement = true;
294 break;
295 case 'd':
296 debugging = true;
297 break;
298 case 'e': /* fewer errors from dodgy lines */
299 iocccok = true;
300 break;
301 case 'f': /* definitions file */
302 defundefile(optarg);
303 break;
304 case 'h':
305 help();
306 break;
307 case 'K': /* keep ambiguous #ifs */
308 strictlogic = true;
309 break;
310 case 'k': /* process constant #ifs */
311 killconsts = true;
312 break;
313 case 'm': /* modify in place */
314 inplace = true;
315 break;
316 case 'M': /* modify in place and keep backup */
317 inplace = true;
318 if (strlen(optarg) > 0)
319 backext = optarg;
320 break;
321 case 'n': /* add #line directive after deleted lines */
322 lnnum = true;
323 break;
324 case 'o': /* output to a file */
325 ofilename = optarg;
326 break;
327 case 's': /* only output list of symbols that control #ifs */
328 symlist = true;
329 break;
330 case 'S': /* list symbols with their nesting depth */
331 symlist = symdepth = true;
332 break;
333 case 't': /* don't parse C comments */
334 text = true;
335 break;
336 case 'V':
337 version();
338 break;
339 case 'x':
340 exitmode = atoi(optarg);
341 if(exitmode < 0 || exitmode > 2)
342 usage();
343 break;
344 default:
345 usage();
346 }
347 argc -= optind;
348 argv += optind;
349 if (compblank && lnblank)
350 errx(2, "-B and -b are mutually exclusive");
351 if (symlist && (ofilename != NULL || inplace || argc > 1))
352 errx(2, "-s only works with one input file");
353 if (argc > 1 && ofilename != NULL)
354 errx(2, "-o cannot be used with multiple input files");
355 if (argc > 1 && !inplace)
356 errx(2, "multiple input files require -m or -M");
357 if (argc == 0 && inplace)
358 errx(2, "-m requires an input file");
359 if (argc == 0)
360 argc = 1;
361 if (argc == 1 && !inplace && ofilename == NULL)
362 ofilename = "-";
363 indirectsym();
364
365 atexit(cleantemp);
366 if (ofilename != NULL)
367 processinout(*argv, ofilename);
368 else while (argc-- > 0) {
369 processinout(*argv, *argv);
370 argv++;
371 }
372 switch(exitmode) {
373 case(0): exit(exitstat);
374 case(1): exit(!exitstat);
375 case(2): exit(0);
376 default: abort(); /* bug */
377 }
378 }
379
380 /*
381 * File logistics.
382 */
383 static void
processinout(const char * ifn,const char * ofn)384 processinout(const char *ifn, const char *ofn)
385 {
386 struct stat st;
387
388 if (ifn == NULL || strcmp(ifn, "-") == 0) {
389 filename = "[stdin]";
390 linefile = NULL;
391 input = fbinmode(stdin);
392 } else {
393 filename = ifn;
394 linefile = ifn;
395 input = fopen(ifn, "rb");
396 if (input == NULL)
397 err(2, "can't open %s", ifn);
398 }
399 if (strcmp(ofn, "-") == 0) {
400 output = fbinmode(stdout);
401 process();
402 return;
403 }
404 if (stat(ofn, &st) < 0) {
405 output = fopen(ofn, "wb");
406 if (output == NULL)
407 err(2, "can't create %s", ofn);
408 process();
409 return;
410 }
411
412 tempname = astrcat(ofn, ".XXXXXX");
413 output = mktempmode(tempname, st.st_mode);
414 if (output == NULL)
415 err(2, "can't create %s", tempname);
416
417 process();
418
419 if (backext != NULL) {
420 char *backname = astrcat(ofn, backext);
421 if (rename(ofn, backname) < 0)
422 err(2, "can't rename \"%s\" to \"%s\"", ofn, backname);
423 free(backname);
424 }
425 /* leave file unmodified if unifdef made no changes */
426 if (!altered && backext == NULL) {
427 if (remove(tempname) < 0)
428 warn("can't remove \"%s\"", tempname);
429 } else if (replace(tempname, ofn) < 0)
430 err(2, "can't rename \"%s\" to \"%s\"", tempname, ofn);
431 free(tempname);
432 tempname = NULL;
433 }
434
435 /*
436 * For cleaning up if there is an error.
437 */
438 static void
cleantemp(void)439 cleantemp(void)
440 {
441 if (tempname != NULL)
442 remove(tempname);
443 }
444
445 /*
446 * Self-identification functions.
447 */
448
449 static void
version(void)450 version(void)
451 {
452 const char *c = copyright;
453 for (;;) {
454 while (*++c != '$')
455 if (*c == '\0')
456 exit(0);
457 while (*++c != '$')
458 putc(*c, stderr);
459 putc('\n', stderr);
460 }
461 }
462
463 static void
synopsis(FILE * fp)464 synopsis(FILE *fp)
465 {
466 fprintf(fp,
467 "usage: unifdef [-bBcdehKkmnsStV] [-x{012}] [-Mext] [-opath] \\\n"
468 " [-[i]Dsym[=val]] [-[i]Usym] [-fpath] ... [file] ...\n");
469 }
470
471 static void
usage(void)472 usage(void)
473 {
474 synopsis(stderr);
475 exit(2);
476 }
477
478 static void
help(void)479 help(void)
480 {
481 synopsis(stdout);
482 printf(
483 " -Dsym=val define preprocessor symbol with given value\n"
484 " -Dsym define preprocessor symbol with value 1\n"
485 " -Usym preprocessor symbol is undefined\n"
486 " -iDsym=val \\ ignore C strings and comments\n"
487 " -iDsym ) in sections controlled by these\n"
488 " -iUsym / preprocessor symbols\n"
489 " -fpath file containing #define and #undef directives\n"
490 " -b blank lines instead of deleting them\n"
491 " -B compress blank lines around deleted section\n"
492 " -c complement (invert) keep vs. delete\n"
493 " -d debugging mode\n"
494 " -e ignore multiline preprocessor directives\n"
495 " -h print help\n"
496 " -Ipath extra include file path (ignored)\n"
497 " -K disable && and || short-circuiting\n"
498 " -k process constant #if expressions\n"
499 " -Mext modify in place and keep backups\n"
500 " -m modify input files in place\n"
501 " -n add #line directives to output\n"
502 " -opath output file name\n"
503 " -S list #if control symbols with nesting\n"
504 " -s list #if control symbols\n"
505 " -t ignore C strings and comments\n"
506 " -V print version\n"
507 " -x{012} exit status mode\n"
508 );
509 exit(0);
510 }
511
512 /*
513 * A state transition function alters the global #if processing state
514 * in a particular way. The table below is indexed by the current
515 * processing state and the type of the current line.
516 *
517 * Nesting is handled by keeping a stack of states; some transition
518 * functions increase or decrease the depth. They also maintain the
519 * ignore state on a stack. In some complicated cases they have to
520 * alter the preprocessor directive, as follows.
521 *
522 * When we have processed a group that starts off with a known-false
523 * #if/#elif sequence (which has therefore been deleted) followed by a
524 * #elif that we don't understand and therefore must keep, we edit the
525 * latter into a #if to keep the nesting correct. We use memcpy() to
526 * overwrite the 4 byte token "elif" with "if " without a '\0' byte.
527 *
528 * When we find a true #elif in a group, the following block will
529 * always be kept and the rest of the sequence after the next #elif or
530 * #else will be discarded. We edit the #elif into a #else and the
531 * following directive to #endif since this has the desired behaviour.
532 *
533 * "Dodgy" directives are split across multiple lines, the most common
534 * example being a multi-line comment hanging off the right of the
535 * directive. We can handle them correctly only if there is no change
536 * from printing to dropping (or vice versa) caused by that directive.
537 * If the directive is the first of a group we have a choice between
538 * failing with an error, or passing it through unchanged instead of
539 * evaluating it. The latter is not the default to avoid questions from
540 * users about unifdef unexpectedly leaving behind preprocessor directives.
541 */
542 typedef void state_fn(void);
543
544 /* report an error */
Eelif(void)545 static void Eelif (void) { error("Inappropriate #elif"); }
Eelse(void)546 static void Eelse (void) { error("Inappropriate #else"); }
Eendif(void)547 static void Eendif(void) { error("Inappropriate #endif"); }
Eeof(void)548 static void Eeof (void) { error("Premature EOF"); }
Eioccc(void)549 static void Eioccc(void) { error("Obfuscated preprocessor control line"); }
550 /* plain line handling */
print(void)551 static void print (void) { flushline(true); }
drop(void)552 static void drop (void) { flushline(false); }
553 /* output lacks group's start line */
Strue(void)554 static void Strue (void) { drop(); ignoreoff(); state(IS_TRUE_PREFIX); }
Sfalse(void)555 static void Sfalse(void) { drop(); ignoreoff(); state(IS_FALSE_PREFIX); }
Selse(void)556 static void Selse (void) { drop(); state(IS_TRUE_ELSE); }
557 /* print/pass this block */
Pelif(void)558 static void Pelif (void) { print(); ignoreoff(); state(IS_PASS_MIDDLE); }
Pelse(void)559 static void Pelse (void) { print(); state(IS_PASS_ELSE); }
Pendif(void)560 static void Pendif(void) { print(); unnest(); }
561 /* discard this block */
Dfalse(void)562 static void Dfalse(void) { drop(); ignoreoff(); state(IS_FALSE_TRAILER); }
Delif(void)563 static void Delif (void) { drop(); ignoreoff(); state(IS_FALSE_MIDDLE); }
Delse(void)564 static void Delse (void) { drop(); state(IS_FALSE_ELSE); }
Dendif(void)565 static void Dendif(void) { drop(); unnest(); }
566 /* first line of group */
Fdrop(void)567 static void Fdrop (void) { nest(); Dfalse(); }
Fpass(void)568 static void Fpass (void) { nest(); Pelif(); }
Ftrue(void)569 static void Ftrue (void) { nest(); Strue(); }
Ffalse(void)570 static void Ffalse(void) { nest(); Sfalse(); }
571 /* variable pedantry for obfuscated lines */
Oiffy(void)572 static void Oiffy (void) { if (!iocccok) Eioccc(); Fpass(); ignoreon(); }
Oif(void)573 static void Oif (void) { if (!iocccok) Eioccc(); Fpass(); }
Oelif(void)574 static void Oelif (void) { if (!iocccok) Eioccc(); Pelif(); }
575 /* ignore comments in this block */
Idrop(void)576 static void Idrop (void) { Fdrop(); ignoreon(); }
Itrue(void)577 static void Itrue (void) { Ftrue(); ignoreon(); }
Ifalse(void)578 static void Ifalse(void) { Ffalse(); ignoreon(); }
579 /* modify this line */
Mpass(void)580 static void Mpass (void) { memcpy(keyword, "if ", 4); Pelif(); }
Mtrue(void)581 static void Mtrue (void) { keywordedit("else"); state(IS_TRUE_MIDDLE); }
Melif(void)582 static void Melif (void) { keywordedit("endif"); state(IS_FALSE_TRAILER); }
Melse(void)583 static void Melse (void) { keywordedit("endif"); state(IS_FALSE_ELSE); }
584
585 static state_fn * const trans_table[IS_COUNT][LT_COUNT] = {
586 /* IS_OUTSIDE */
587 { Itrue, Ifalse,Fpass, Ftrue, Ffalse,Eelif, Eelif, Eelif, Eelse, Eendif,
588 Oiffy, Oiffy, Fpass, Oif, Oif, Eelif, Eelif, Eelif, Eelse, Eendif,
589 print, done, abort },
590 /* IS_FALSE_PREFIX */
591 { Idrop, Idrop, Fdrop, Fdrop, Fdrop, Mpass, Strue, Sfalse,Selse, Dendif,
592 Idrop, Idrop, Fdrop, Fdrop, Fdrop, Mpass, Eioccc,Eioccc,Eioccc,Eioccc,
593 drop, Eeof, abort },
594 /* IS_TRUE_PREFIX */
595 { Itrue, Ifalse,Fpass, Ftrue, Ffalse,Dfalse,Dfalse,Dfalse,Delse, Dendif,
596 Oiffy, Oiffy, Fpass, Oif, Oif, Eioccc,Eioccc,Eioccc,Eioccc,Eioccc,
597 print, Eeof, abort },
598 /* IS_PASS_MIDDLE */
599 { Itrue, Ifalse,Fpass, Ftrue, Ffalse,Pelif, Mtrue, Delif, Pelse, Pendif,
600 Oiffy, Oiffy, Fpass, Oif, Oif, Pelif, Oelif, Oelif, Pelse, Pendif,
601 print, Eeof, abort },
602 /* IS_FALSE_MIDDLE */
603 { Idrop, Idrop, Fdrop, Fdrop, Fdrop, Pelif, Mtrue, Delif, Pelse, Pendif,
604 Idrop, Idrop, Fdrop, Fdrop, Fdrop, Eioccc,Eioccc,Eioccc,Eioccc,Eioccc,
605 drop, Eeof, abort },
606 /* IS_TRUE_MIDDLE */
607 { Itrue, Ifalse,Fpass, Ftrue, Ffalse,Melif, Melif, Melif, Melse, Pendif,
608 Oiffy, Oiffy, Fpass, Oif, Oif, Eioccc,Eioccc,Eioccc,Eioccc,Pendif,
609 print, Eeof, abort },
610 /* IS_PASS_ELSE */
611 { Itrue, Ifalse,Fpass, Ftrue, Ffalse,Eelif, Eelif, Eelif, Eelse, Pendif,
612 Oiffy, Oiffy, Fpass, Oif, Oif, Eelif, Eelif, Eelif, Eelse, Pendif,
613 print, Eeof, abort },
614 /* IS_FALSE_ELSE */
615 { Idrop, Idrop, Fdrop, Fdrop, Fdrop, Eelif, Eelif, Eelif, Eelse, Dendif,
616 Idrop, Idrop, Fdrop, Fdrop, Fdrop, Eelif, Eelif, Eelif, Eelse, Eioccc,
617 drop, Eeof, abort },
618 /* IS_TRUE_ELSE */
619 { Itrue, Ifalse,Fpass, Ftrue, Ffalse,Eelif, Eelif, Eelif, Eelse, Dendif,
620 Oiffy, Oiffy, Fpass, Oif, Oif, Eelif, Eelif, Eelif, Eelse, Eioccc,
621 print, Eeof, abort },
622 /* IS_FALSE_TRAILER */
623 { Idrop, Idrop, Fdrop, Fdrop, Fdrop, Dfalse,Dfalse,Dfalse,Delse, Dendif,
624 Idrop, Idrop, Fdrop, Fdrop, Fdrop, Dfalse,Dfalse,Dfalse,Delse, Eioccc,
625 drop, Eeof, abort }
626 /*TRUEI FALSEI IF TRUE FALSE ELIF ELTRUE ELFALSE ELSE ENDIF
627 TRUEI FALSEI IF TRUE FALSE ELIF ELTRUE ELFALSE ELSE ENDIF (DODGY)
628 PLAIN EOF ERROR */
629 };
630
631 /*
632 * State machine utility functions
633 */
634 static void
ignoreoff(void)635 ignoreoff(void)
636 {
637 if (depth == 0)
638 abort(); /* bug */
639 ignoring[depth] = ignoring[depth-1];
640 }
641 static void
ignoreon(void)642 ignoreon(void)
643 {
644 ignoring[depth] = true;
645 }
646 static void
keywordedit(const char * replacement)647 keywordedit(const char *replacement)
648 {
649 snprintf(keyword, tline + sizeof(tline) - keyword,
650 "%s%s", replacement, newline);
651 altered = true;
652 print();
653 }
654 static void
nest(void)655 nest(void)
656 {
657 if (depth > MAXDEPTH-1)
658 abort(); /* bug */
659 if (depth == MAXDEPTH-1)
660 error("Too many levels of nesting");
661 depth += 1;
662 stifline[depth] = linenum;
663 }
664 static void
unnest(void)665 unnest(void)
666 {
667 if (depth == 0)
668 abort(); /* bug */
669 depth -= 1;
670 }
671 static void
state(Ifstate is)672 state(Ifstate is)
673 {
674 ifstate[depth] = is;
675 }
676
677 /*
678 * The last state transition function. When this is called,
679 * lineval == LT_EOF, so the process() loop will terminate.
680 */
681 static void
done(void)682 done(void)
683 {
684 if (incomment)
685 error("EOF in comment");
686 closeio();
687 }
688
689 /*
690 * Write a line to the output or not, according to command line options.
691 * If writing fails, closeio() will print the error and exit.
692 */
693 static void
flushline(bool keep)694 flushline(bool keep)
695 {
696 if (symlist)
697 return;
698 if (keep ^ complement) {
699 bool blankline = tline[strspn(tline, " \t\r\n")] == '\0';
700 if (blankline && compblank && blankcount != blankmax) {
701 delcount += 1;
702 blankcount += 1;
703 } else {
704 if (lnnum && delcount > 0)
705 hashline();
706 if (fputs(tline, output) == EOF)
707 closeio();
708 delcount = 0;
709 blankmax = blankcount = blankline ? blankcount + 1 : 0;
710 }
711 } else {
712 if (lnblank && fputs(newline, output) == EOF)
713 closeio();
714 altered = true;
715 delcount += 1;
716 blankcount = 0;
717 }
718 if (debugging && fflush(output) == EOF)
719 closeio();
720 }
721
722 /*
723 * Format of #line directives depends on whether we know the input filename.
724 */
725 static void
hashline(void)726 hashline(void)
727 {
728 int e;
729
730 if (linefile == NULL)
731 e = fprintf(output, "#line %d%s", linenum, newline);
732 else
733 e = fprintf(output, "#line %d \"%s\"%s",
734 linenum, linefile, newline);
735 if (e < 0)
736 closeio();
737 }
738
739 /*
740 * Flush the output and handle errors.
741 */
742 static void
closeio(void)743 closeio(void)
744 {
745 /* Tidy up after findsym(). */
746 if (symdepth && !zerosyms)
747 printf("\n");
748 if (output != NULL && (ferror(output) || fclose(output) == EOF))
749 err(2, "%s: can't write to output", filename);
750 fclose(input);
751 }
752
753 /*
754 * The driver for the state machine.
755 */
756 static void
process(void)757 process(void)
758 {
759 Linetype lineval = LT_PLAIN;
760 /* When compressing blank lines, act as if the file
761 is preceded by a large number of blank lines. */
762 blankmax = blankcount = 1000;
763 zerosyms = true;
764 newline = NULL;
765 linenum = 0;
766 altered = false;
767 while (lineval != LT_EOF) {
768 lineval = parseline();
769 trans_table[ifstate[depth]][lineval]();
770 debug("process line %d %s -> %s depth %d",
771 linenum, linetype_name[lineval],
772 ifstate_name[ifstate[depth]], depth);
773 }
774 exitstat |= altered;
775 }
776
777 /*
778 * Parse a line and determine its type. We keep the preprocessor line
779 * parser state between calls in the global variable linestate, with
780 * help from skipcomment().
781 */
782 static Linetype
parseline(void)783 parseline(void)
784 {
785 const char *cp;
786 int cursym;
787 Linetype retval;
788 Comment_state wascomment;
789
790 wascomment = incomment;
791 cp = skiphash();
792 if (cp == NULL)
793 return (LT_EOF);
794 if (newline == NULL) {
795 if (strrchr(tline, '\n') == strrchr(tline, '\r') + 1)
796 newline = newline_crlf;
797 else
798 newline = newline_unix;
799 }
800 if (*cp == '\0') {
801 retval = LT_PLAIN;
802 goto done;
803 }
804 keyword = tline + (cp - tline);
805 if ((cp = matchsym("ifdef", keyword)) != NULL ||
806 (cp = matchsym("ifndef", keyword)) != NULL) {
807 cp = skipcomment(cp);
808 if ((cursym = findsym(&cp)) < 0)
809 retval = LT_IF;
810 else {
811 retval = (keyword[2] == 'n')
812 ? LT_FALSE : LT_TRUE;
813 if (value[cursym] == NULL)
814 retval = (retval == LT_TRUE)
815 ? LT_FALSE : LT_TRUE;
816 if (ignore[cursym])
817 retval = (retval == LT_TRUE)
818 ? LT_TRUEI : LT_FALSEI;
819 }
820 } else if ((cp = matchsym("if", keyword)) != NULL)
821 retval = ifeval(&cp);
822 else if ((cp = matchsym("elif", keyword)) != NULL)
823 retval = linetype_if2elif(ifeval(&cp));
824 else if ((cp = matchsym("else", keyword)) != NULL)
825 retval = LT_ELSE;
826 else if ((cp = matchsym("endif", keyword)) != NULL)
827 retval = LT_ENDIF;
828 else {
829 cp = skipsym(keyword);
830 /* no way can we deal with a continuation inside a keyword */
831 if (strncmp(cp, "\\\r\n", 3) == 0 ||
832 strncmp(cp, "\\\n", 2) == 0)
833 Eioccc();
834 cp = skipline(cp);
835 retval = LT_PLAIN;
836 goto done;
837 }
838 cp = skipcomment(cp);
839 if (*cp != '\0') {
840 cp = skipline(cp);
841 if (retval == LT_TRUE || retval == LT_FALSE ||
842 retval == LT_TRUEI || retval == LT_FALSEI)
843 retval = LT_IF;
844 if (retval == LT_ELTRUE || retval == LT_ELFALSE)
845 retval = LT_ELIF;
846 }
847 /* the following can happen if the last line of the file lacks a
848 newline or if there is too much whitespace in a directive */
849 if (linestate == LS_HASH) {
850 long len = cp - tline;
851 if (fgets(tline + len, MAXLINE - len, input) == NULL) {
852 if (ferror(input))
853 err(2, "can't read %s", filename);
854 /* append the missing newline at eof */
855 strcpy(tline + len, newline);
856 cp += strlen(newline);
857 linestate = LS_START;
858 } else {
859 linestate = LS_DIRTY;
860 }
861 }
862 if (retval != LT_PLAIN && (wascomment || linestate != LS_START)) {
863 retval = linetype_2dodgy(retval);
864 linestate = LS_DIRTY;
865 }
866 done:
867 debug("parser line %d state %s comment %s line", linenum,
868 comment_name[incomment], linestate_name[linestate]);
869 return (retval);
870 }
871
872 /*
873 * These are the binary operators that are supported by the expression
874 * evaluator.
875 */
op_strict(long * p,long v,Linetype at,Linetype bt)876 static Linetype op_strict(long *p, long v, Linetype at, Linetype bt) {
877 if(at == LT_IF || bt == LT_IF) return (LT_IF);
878 return (*p = v, v ? LT_TRUE : LT_FALSE);
879 }
op_lt(long * p,Linetype at,long a,Linetype bt,long b)880 static Linetype op_lt(long *p, Linetype at, long a, Linetype bt, long b) {
881 return op_strict(p, a < b, at, bt);
882 }
op_gt(long * p,Linetype at,long a,Linetype bt,long b)883 static Linetype op_gt(long *p, Linetype at, long a, Linetype bt, long b) {
884 return op_strict(p, a > b, at, bt);
885 }
op_le(long * p,Linetype at,long a,Linetype bt,long b)886 static Linetype op_le(long *p, Linetype at, long a, Linetype bt, long b) {
887 return op_strict(p, a <= b, at, bt);
888 }
op_ge(long * p,Linetype at,long a,Linetype bt,long b)889 static Linetype op_ge(long *p, Linetype at, long a, Linetype bt, long b) {
890 return op_strict(p, a >= b, at, bt);
891 }
op_eq(long * p,Linetype at,long a,Linetype bt,long b)892 static Linetype op_eq(long *p, Linetype at, long a, Linetype bt, long b) {
893 return op_strict(p, a == b, at, bt);
894 }
op_ne(long * p,Linetype at,long a,Linetype bt,long b)895 static Linetype op_ne(long *p, Linetype at, long a, Linetype bt, long b) {
896 return op_strict(p, a != b, at, bt);
897 }
op_or(long * p,Linetype at,long a,Linetype bt,long b)898 static Linetype op_or(long *p, Linetype at, long a, Linetype bt, long b) {
899 if (!strictlogic && (at == LT_TRUE || bt == LT_TRUE))
900 return (*p = 1, LT_TRUE);
901 return op_strict(p, a || b, at, bt);
902 }
op_and(long * p,Linetype at,long a,Linetype bt,long b)903 static Linetype op_and(long *p, Linetype at, long a, Linetype bt, long b) {
904 if (!strictlogic && (at == LT_FALSE || bt == LT_FALSE))
905 return (*p = 0, LT_FALSE);
906 return op_strict(p, a && b, at, bt);
907 }
op_blsh(long * p,Linetype at,long a,Linetype bt,long b)908 static Linetype op_blsh(long *p, Linetype at, long a, Linetype bt, long b) {
909 return op_strict(p, a << b, at, bt);
910 }
op_brsh(long * p,Linetype at,long a,Linetype bt,long b)911 static Linetype op_brsh(long *p, Linetype at, long a, Linetype bt, long b) {
912 return op_strict(p, a >> b, at, bt);
913 }
op_add(long * p,Linetype at,long a,Linetype bt,long b)914 static Linetype op_add(long *p, Linetype at, long a, Linetype bt, long b) {
915 return op_strict(p, a + b, at, bt);
916 }
op_sub(long * p,Linetype at,long a,Linetype bt,long b)917 static Linetype op_sub(long *p, Linetype at, long a, Linetype bt, long b) {
918 return op_strict(p, a - b, at, bt);
919 }
op_mul(long * p,Linetype at,long a,Linetype bt,long b)920 static Linetype op_mul(long *p, Linetype at, long a, Linetype bt, long b) {
921 return op_strict(p, a * b, at, bt);
922 }
op_div(long * p,Linetype at,long a,Linetype bt,long b)923 static Linetype op_div(long *p, Linetype at, long a, Linetype bt, long b) {
924 if (bt != LT_TRUE) {
925 debug("eval division by zero");
926 return (LT_ERROR);
927 }
928 return op_strict(p, a / b, at, bt);
929 }
op_mod(long * p,Linetype at,long a,Linetype bt,long b)930 static Linetype op_mod(long *p, Linetype at, long a, Linetype bt, long b) {
931 return op_strict(p, a % b, at, bt);
932 }
op_bor(long * p,Linetype at,long a,Linetype bt,long b)933 static Linetype op_bor(long *p, Linetype at, long a, Linetype bt, long b) {
934 return op_strict(p, a | b, at, bt);
935 }
op_bxor(long * p,Linetype at,long a,Linetype bt,long b)936 static Linetype op_bxor(long *p, Linetype at, long a, Linetype bt, long b) {
937 return op_strict(p, a ^ b, at, bt);
938 }
op_band(long * p,Linetype at,long a,Linetype bt,long b)939 static Linetype op_band(long *p, Linetype at, long a, Linetype bt, long b) {
940 return op_strict(p, a & b, at, bt);
941 }
942
943 /*
944 * An evaluation function takes three arguments, as follows: (1) a pointer to
945 * an element of the precedence table which lists the operators at the current
946 * level of precedence; (2) a pointer to an integer which will receive the
947 * value of the expression; and (3) a pointer to a char* that points to the
948 * expression to be evaluated and that is updated to the end of the expression
949 * when evaluation is complete. The function returns LT_FALSE if the value of
950 * the expression is zero, LT_TRUE if it is non-zero, LT_IF if the expression
951 * depends on an unknown symbol, or LT_ERROR if there is a parse failure.
952 */
953 struct ops;
954
955 typedef Linetype eval_fn(const struct ops *, long *, const char **);
956
957 static eval_fn eval_table, eval_unary;
958
959 /*
960 * The precedence table. Expressions involving binary operators are evaluated
961 * in a table-driven way by eval_table. When it evaluates a subexpression it
962 * calls the inner function with its first argument pointing to the next
963 * element of the table. Innermost expressions have special non-table-driven
964 * handling.
965 *
966 * The stop characters help with lexical analysis: an operator is not
967 * recognized if it is followed by one of the stop characters because
968 * that would make it a different operator.
969 */
970 struct op {
971 const char *str;
972 Linetype (*fn)(long *, Linetype, long, Linetype, long);
973 const char *stop;
974 };
975 struct ops {
976 eval_fn *inner;
977 struct op op[5];
978 };
979 static const struct ops eval_ops[] = {
980 { eval_table, { { "||", op_or, NULL } } },
981 { eval_table, { { "&&", op_and, NULL } } },
982 { eval_table, { { "|", op_bor, "|" } } },
983 { eval_table, { { "^", op_bxor, NULL } } },
984 { eval_table, { { "&", op_band, "&" } } },
985 { eval_table, { { "==", op_eq, NULL },
986 { "!=", op_ne, NULL } } },
987 { eval_table, { { "<=", op_le, NULL },
988 { ">=", op_ge, NULL },
989 { "<", op_lt, "<=" },
990 { ">", op_gt, ">=" } } },
991 { eval_table, { { "<<", op_blsh, NULL },
992 { ">>", op_brsh, NULL } } },
993 { eval_table, { { "+", op_add, NULL },
994 { "-", op_sub, NULL } } },
995 { eval_unary, { { "*", op_mul, NULL },
996 { "/", op_div, NULL },
997 { "%", op_mod, NULL } } },
998 };
999
1000 /* Current operator precedence level */
prec(const struct ops * ops)1001 static long prec(const struct ops *ops)
1002 {
1003 return (ops - eval_ops);
1004 }
1005
1006 /*
1007 * Function for evaluating the innermost parts of expressions,
1008 * viz. !expr (expr) number defined(symbol) symbol
1009 * We reset the constexpr flag in the last two cases.
1010 */
1011 static Linetype
eval_unary(const struct ops * ops,long * valp,const char ** cpp)1012 eval_unary(const struct ops *ops, long *valp, const char **cpp)
1013 {
1014 const char *cp;
1015 char *ep;
1016 int sym;
1017 bool defparen;
1018 Linetype lt;
1019
1020 cp = skipcomment(*cpp);
1021 if (*cp == '!') {
1022 debug("eval%d !", prec(ops));
1023 cp++;
1024 lt = eval_unary(ops, valp, &cp);
1025 if (lt == LT_ERROR)
1026 return (LT_ERROR);
1027 if (lt != LT_IF) {
1028 *valp = !*valp;
1029 lt = *valp ? LT_TRUE : LT_FALSE;
1030 }
1031 } else if (*cp == '~') {
1032 debug("eval%d ~", prec(ops));
1033 cp++;
1034 lt = eval_unary(ops, valp, &cp);
1035 if (lt == LT_ERROR)
1036 return (LT_ERROR);
1037 if (lt != LT_IF) {
1038 *valp = ~(*valp);
1039 lt = *valp ? LT_TRUE : LT_FALSE;
1040 }
1041 } else if (*cp == '-') {
1042 debug("eval%d -", prec(ops));
1043 cp++;
1044 lt = eval_unary(ops, valp, &cp);
1045 if (lt == LT_ERROR)
1046 return (LT_ERROR);
1047 if (lt != LT_IF) {
1048 *valp = -(*valp);
1049 lt = *valp ? LT_TRUE : LT_FALSE;
1050 }
1051 } else if (*cp == '(') {
1052 cp++;
1053 debug("eval%d (", prec(ops));
1054 lt = eval_table(eval_ops, valp, &cp);
1055 if (lt == LT_ERROR)
1056 return (LT_ERROR);
1057 cp = skipcomment(cp);
1058 if (*cp++ != ')')
1059 return (LT_ERROR);
1060 } else if (isdigit((unsigned char)*cp)) {
1061 debug("eval%d number", prec(ops));
1062 *valp = strtol(cp, &ep, 0);
1063 if (ep == cp)
1064 return (LT_ERROR);
1065 lt = *valp ? LT_TRUE : LT_FALSE;
1066 cp = ep;
1067 } else if (matchsym("defined", cp) != NULL) {
1068 cp = skipcomment(cp+7);
1069 if (*cp == '(') {
1070 cp = skipcomment(cp+1);
1071 defparen = true;
1072 } else {
1073 defparen = false;
1074 }
1075 sym = findsym(&cp);
1076 cp = skipcomment(cp);
1077 if (defparen && *cp++ != ')') {
1078 debug("eval%d defined missing ')'", prec(ops));
1079 return (LT_ERROR);
1080 }
1081 if (sym < 0) {
1082 debug("eval%d defined unknown", prec(ops));
1083 lt = LT_IF;
1084 } else {
1085 debug("eval%d defined %s", prec(ops), symname[sym]);
1086 *valp = (value[sym] != NULL);
1087 lt = *valp ? LT_TRUE : LT_FALSE;
1088 }
1089 constexpr = false;
1090 } else if (!endsym(*cp)) {
1091 debug("eval%d symbol", prec(ops));
1092 sym = findsym(&cp);
1093 if (sym < 0) {
1094 lt = LT_IF;
1095 cp = skipargs(cp);
1096 } else if (value[sym] == NULL) {
1097 *valp = 0;
1098 lt = LT_FALSE;
1099 } else {
1100 *valp = strtol(value[sym], &ep, 0);
1101 if (*ep != '\0' || ep == value[sym])
1102 return (LT_ERROR);
1103 lt = *valp ? LT_TRUE : LT_FALSE;
1104 cp = skipargs(cp);
1105 }
1106 constexpr = false;
1107 } else {
1108 debug("eval%d bad expr", prec(ops));
1109 return (LT_ERROR);
1110 }
1111
1112 *cpp = cp;
1113 debug("eval%d = %d", prec(ops), *valp);
1114 return (lt);
1115 }
1116
1117 /*
1118 * Table-driven evaluation of binary operators.
1119 */
1120 static Linetype
eval_table(const struct ops * ops,long * valp,const char ** cpp)1121 eval_table(const struct ops *ops, long *valp, const char **cpp)
1122 {
1123 const struct op *op;
1124 const char *cp;
1125 long val = 0;
1126 Linetype lt, rt;
1127
1128 debug("eval%d", prec(ops));
1129 cp = *cpp;
1130 lt = ops->inner(ops+1, valp, &cp);
1131 if (lt == LT_ERROR)
1132 return (LT_ERROR);
1133 for (;;) {
1134 cp = skipcomment(cp);
1135 for (op = ops->op; op->str != NULL; op++) {
1136 if (strncmp(cp, op->str, strlen(op->str)) == 0) {
1137 /* assume only one-char operators have stop chars */
1138 if (op->stop != NULL && cp[1] != '\0' &&
1139 strchr(op->stop, cp[1]) != NULL)
1140 continue;
1141 else
1142 break;
1143 }
1144 }
1145 if (op->str == NULL)
1146 break;
1147 cp += strlen(op->str);
1148 debug("eval%d %s", prec(ops), op->str);
1149 rt = ops->inner(ops+1, &val, &cp);
1150 if (rt == LT_ERROR)
1151 return (LT_ERROR);
1152 lt = op->fn(valp, lt, *valp, rt, val);
1153 }
1154
1155 *cpp = cp;
1156 debug("eval%d = %d", prec(ops), *valp);
1157 debug("eval%d lt = %s", prec(ops), linetype_name[lt]);
1158 return (lt);
1159 }
1160
1161 /*
1162 * Evaluate the expression on a #if or #elif line. If we can work out
1163 * the result we return LT_TRUE or LT_FALSE accordingly, otherwise we
1164 * return just a generic LT_IF.
1165 */
1166 static Linetype
ifeval(const char ** cpp)1167 ifeval(const char **cpp)
1168 {
1169 Linetype ret;
1170 long val = 0;
1171
1172 debug("eval %s", *cpp);
1173 constexpr = killconsts ? false : true;
1174 ret = eval_table(eval_ops, &val, cpp);
1175 debug("eval = %d", val);
1176 return (constexpr ? LT_IF : ret == LT_ERROR ? LT_IF : ret);
1177 }
1178
1179 /*
1180 * Read a line and examine its initial part to determine if it is a
1181 * preprocessor directive. Returns NULL on EOF, or a pointer to a
1182 * preprocessor directive name, or a pointer to the zero byte at the
1183 * end of the line.
1184 */
1185 static const char *
skiphash(void)1186 skiphash(void)
1187 {
1188 const char *cp;
1189
1190 linenum++;
1191 if (fgets(tline, MAXLINE, input) == NULL) {
1192 if (ferror(input))
1193 err(2, "can't read %s", filename);
1194 else
1195 return (NULL);
1196 }
1197 cp = skipcomment(tline);
1198 if (linestate == LS_START && *cp == '#') {
1199 linestate = LS_HASH;
1200 return (skipcomment(cp + 1));
1201 } else if (*cp == '\0') {
1202 return (cp);
1203 } else {
1204 return (skipline(cp));
1205 }
1206 }
1207
1208 /*
1209 * Mark a line dirty and consume the rest of it, keeping track of the
1210 * lexical state.
1211 */
1212 static const char *
skipline(const char * cp)1213 skipline(const char *cp)
1214 {
1215 const char *pcp;
1216 if (*cp != '\0')
1217 linestate = LS_DIRTY;
1218 while (*cp != '\0') {
1219 cp = skipcomment(pcp = cp);
1220 if (pcp == cp)
1221 cp++;
1222 }
1223 return (cp);
1224 }
1225
1226 /*
1227 * Skip over comments, strings, and character literals and stop at the
1228 * next character position that is not whitespace. Between calls we keep
1229 * the comment state in the global variable incomment, and we also adjust
1230 * the global variable linestate when we see a newline.
1231 * XXX: doesn't cope with the buffer splitting inside a state transition.
1232 */
1233 static const char *
skipcomment(const char * cp)1234 skipcomment(const char *cp)
1235 {
1236 if (text || ignoring[depth]) {
1237 for (; isspace((unsigned char)*cp); cp++)
1238 if (*cp == '\n')
1239 linestate = LS_START;
1240 return (cp);
1241 }
1242 while (*cp != '\0')
1243 /* don't reset to LS_START after a line continuation */
1244 if (strncmp(cp, "\\\r\n", 3) == 0)
1245 cp += 3;
1246 else if (strncmp(cp, "\\\n", 2) == 0)
1247 cp += 2;
1248 else switch (incomment) {
1249 case NO_COMMENT:
1250 if (strncmp(cp, "/\\\r\n", 4) == 0) {
1251 incomment = STARTING_COMMENT;
1252 cp += 4;
1253 } else if (strncmp(cp, "/\\\n", 3) == 0) {
1254 incomment = STARTING_COMMENT;
1255 cp += 3;
1256 } else if (strncmp(cp, "/*", 2) == 0) {
1257 incomment = C_COMMENT;
1258 cp += 2;
1259 } else if (strncmp(cp, "//", 2) == 0) {
1260 incomment = CXX_COMMENT;
1261 cp += 2;
1262 } else if (strncmp(cp, "\'", 1) == 0) {
1263 incomment = CHAR_LITERAL;
1264 linestate = LS_DIRTY;
1265 cp += 1;
1266 } else if (strncmp(cp, "\"", 1) == 0) {
1267 incomment = STRING_LITERAL;
1268 linestate = LS_DIRTY;
1269 cp += 1;
1270 } else if (strncmp(cp, "\n", 1) == 0) {
1271 linestate = LS_START;
1272 cp += 1;
1273 } else if (strchr(" \r\t", *cp) != NULL) {
1274 cp += 1;
1275 } else
1276 return (cp);
1277 continue;
1278 case CXX_COMMENT:
1279 if (strncmp(cp, "\n", 1) == 0) {
1280 incomment = NO_COMMENT;
1281 linestate = LS_START;
1282 }
1283 cp += 1;
1284 continue;
1285 case CHAR_LITERAL:
1286 case STRING_LITERAL:
1287 if ((incomment == CHAR_LITERAL && cp[0] == '\'') ||
1288 (incomment == STRING_LITERAL && cp[0] == '\"')) {
1289 incomment = NO_COMMENT;
1290 cp += 1;
1291 } else if (cp[0] == '\\') {
1292 if (cp[1] == '\0')
1293 cp += 1;
1294 else
1295 cp += 2;
1296 } else if (strncmp(cp, "\n", 1) == 0) {
1297 if (incomment == CHAR_LITERAL)
1298 error("Unterminated char literal");
1299 else
1300 error("Unterminated string literal");
1301 } else
1302 cp += 1;
1303 continue;
1304 case C_COMMENT:
1305 if (strncmp(cp, "*\\\r\n", 4) == 0) {
1306 incomment = FINISHING_COMMENT;
1307 cp += 4;
1308 } else if (strncmp(cp, "*\\\n", 3) == 0) {
1309 incomment = FINISHING_COMMENT;
1310 cp += 3;
1311 } else if (strncmp(cp, "*/", 2) == 0) {
1312 incomment = NO_COMMENT;
1313 cp += 2;
1314 } else
1315 cp += 1;
1316 continue;
1317 case STARTING_COMMENT:
1318 if (*cp == '*') {
1319 incomment = C_COMMENT;
1320 cp += 1;
1321 } else if (*cp == '/') {
1322 incomment = CXX_COMMENT;
1323 cp += 1;
1324 } else {
1325 incomment = NO_COMMENT;
1326 linestate = LS_DIRTY;
1327 }
1328 continue;
1329 case FINISHING_COMMENT:
1330 if (*cp == '/') {
1331 incomment = NO_COMMENT;
1332 cp += 1;
1333 } else
1334 incomment = C_COMMENT;
1335 continue;
1336 default:
1337 abort(); /* bug */
1338 }
1339 return (cp);
1340 }
1341
1342 /*
1343 * Skip macro arguments.
1344 */
1345 static const char *
skipargs(const char * cp)1346 skipargs(const char *cp)
1347 {
1348 const char *ocp = cp;
1349 int level = 0;
1350 cp = skipcomment(cp);
1351 if (*cp != '(')
1352 return (cp);
1353 do {
1354 if (*cp == '(')
1355 level++;
1356 if (*cp == ')')
1357 level--;
1358 cp = skipcomment(cp+1);
1359 } while (level != 0 && *cp != '\0');
1360 if (level == 0)
1361 return (cp);
1362 else
1363 /* Rewind and re-detect the syntax error later. */
1364 return (ocp);
1365 }
1366
1367 /*
1368 * Skip over an identifier.
1369 */
1370 static const char *
skipsym(const char * cp)1371 skipsym(const char *cp)
1372 {
1373 while (!endsym(*cp))
1374 ++cp;
1375 return (cp);
1376 }
1377
1378 /*
1379 * Skip whitespace and take a copy of any following identifier.
1380 */
1381 static const char *
getsym(const char ** cpp)1382 getsym(const char **cpp)
1383 {
1384 const char *cp = *cpp, *sym;
1385
1386 cp = skipcomment(cp);
1387 cp = skipsym(sym = cp);
1388 if (cp == sym)
1389 return NULL;
1390 *cpp = cp;
1391 return (xstrdup(sym, cp));
1392 }
1393
1394 /*
1395 * Check that s (a symbol) matches the start of t, and that the
1396 * following character in t is not a symbol character. Returns a
1397 * pointer to the following character in t if there is a match,
1398 * otherwise NULL.
1399 */
1400 static const char *
matchsym(const char * s,const char * t)1401 matchsym(const char *s, const char *t)
1402 {
1403 while (*s != '\0' && *t != '\0')
1404 if (*s != *t)
1405 return (NULL);
1406 else
1407 ++s, ++t;
1408 if (*s == '\0' && endsym(*t))
1409 return(t);
1410 else
1411 return(NULL);
1412 }
1413
1414 /*
1415 * Look for the symbol in the symbol table. If it is found, we return
1416 * the symbol table index, else we return -1.
1417 */
1418 static int
findsym(const char ** strp)1419 findsym(const char **strp)
1420 {
1421 const char *str;
1422 int symind;
1423
1424 str = *strp;
1425 *strp = skipsym(str);
1426 if (symlist) {
1427 if (*strp == str)
1428 return (-1);
1429 if (symdepth && firstsym)
1430 printf("%s%3d", zerosyms ? "" : "\n", depth);
1431 firstsym = zerosyms = false;
1432 printf("%s%.*s%s",
1433 symdepth ? " " : "",
1434 (int)(*strp-str), str,
1435 symdepth ? "" : "\n");
1436 /* we don't care about the value of the symbol */
1437 return (0);
1438 }
1439 for (symind = 0; symind < nsyms; ++symind) {
1440 if (matchsym(symname[symind], str) != NULL) {
1441 debugsym("findsym", symind);
1442 return (symind);
1443 }
1444 }
1445 return (-1);
1446 }
1447
1448 /*
1449 * Resolve indirect symbol values to their final definitions.
1450 */
1451 static void
indirectsym(void)1452 indirectsym(void)
1453 {
1454 const char *cp;
1455 int changed, sym, ind;
1456
1457 do {
1458 changed = 0;
1459 for (sym = 0; sym < nsyms; ++sym) {
1460 if (value[sym] == NULL)
1461 continue;
1462 cp = value[sym];
1463 ind = findsym(&cp);
1464 if (ind == -1 || ind == sym ||
1465 *cp != '\0' ||
1466 value[ind] == NULL ||
1467 value[ind] == value[sym])
1468 continue;
1469 debugsym("indir...", sym);
1470 value[sym] = value[ind];
1471 debugsym("...ectsym", sym);
1472 changed++;
1473 }
1474 } while (changed);
1475 }
1476
1477 /*
1478 * Add a symbol to the symbol table, specified with the format sym=val
1479 */
1480 static void
addsym1(bool ignorethis,bool definethis,char * symval)1481 addsym1(bool ignorethis, bool definethis, char *symval)
1482 {
1483 const char *sym, *val;
1484
1485 sym = symval;
1486 val = skipsym(sym);
1487 if (definethis && *val == '=') {
1488 symval[val - sym] = '\0';
1489 val = val + 1;
1490 } else if (*val == '\0') {
1491 val = definethis ? "1" : NULL;
1492 } else {
1493 usage();
1494 }
1495 addsym2(ignorethis, sym, val);
1496 }
1497
1498 /*
1499 * Add a symbol to the symbol table.
1500 */
1501 static void
addsym2(bool ignorethis,const char * sym,const char * val)1502 addsym2(bool ignorethis, const char *sym, const char *val)
1503 {
1504 const char *cp = sym;
1505 int symind;
1506
1507 symind = findsym(&cp);
1508 if (symind < 0) {
1509 if (nsyms >= MAXSYMS)
1510 errx(2, "too many symbols");
1511 symind = nsyms++;
1512 }
1513 ignore[symind] = ignorethis;
1514 symname[symind] = sym;
1515 value[symind] = val;
1516 debugsym("addsym", symind);
1517 }
1518
1519 static void
debugsym(const char * why,int symind)1520 debugsym(const char *why, int symind)
1521 {
1522 debug("%s %s%c%s", why, symname[symind],
1523 value[symind] ? '=' : ' ',
1524 value[symind] ? value[symind] : "undef");
1525 }
1526
1527 /*
1528 * Add symbols to the symbol table from a file containing
1529 * #define and #undef preprocessor directives.
1530 */
1531 static void
defundefile(const char * fn)1532 defundefile(const char *fn)
1533 {
1534 filename = fn;
1535 input = fopen(fn, "rb");
1536 if (input == NULL)
1537 err(2, "can't open %s", fn);
1538 linenum = 0;
1539 while (defundef())
1540 ;
1541 if (ferror(input))
1542 err(2, "can't read %s", filename);
1543 else
1544 fclose(input);
1545 if (incomment)
1546 error("EOF in comment");
1547 }
1548
1549 /*
1550 * Read and process one #define or #undef directive
1551 */
1552 static bool
defundef(void)1553 defundef(void)
1554 {
1555 const char *cp, *kw, *sym, *val, *end;
1556
1557 cp = skiphash();
1558 if (cp == NULL)
1559 return (false);
1560 if (*cp == '\0')
1561 goto done;
1562 /* strip trailing whitespace, and do a fairly rough check to
1563 avoid unsupported multi-line preprocessor directives */
1564 end = cp + strlen(cp);
1565 while (end > tline && strchr(" \t\n\r", end[-1]) != NULL)
1566 --end;
1567 if (end > tline && end[-1] == '\\')
1568 Eioccc();
1569
1570 kw = cp;
1571 if ((cp = matchsym("define", kw)) != NULL) {
1572 sym = getsym(&cp);
1573 if (sym == NULL)
1574 error("Missing macro name in #define");
1575 if (*cp == '(') {
1576 val = "1";
1577 } else {
1578 cp = skipcomment(cp);
1579 val = (cp < end) ? xstrdup(cp, end) : "";
1580 }
1581 debug("#define");
1582 addsym2(false, sym, val);
1583 } else if ((cp = matchsym("undef", kw)) != NULL) {
1584 sym = getsym(&cp);
1585 if (sym == NULL)
1586 error("Missing macro name in #undef");
1587 cp = skipcomment(cp);
1588 debug("#undef");
1589 addsym2(false, sym, NULL);
1590 } else {
1591 error("Unrecognized preprocessor directive");
1592 }
1593 skipline(cp);
1594 done:
1595 debug("parser line %d state %s comment %s line", linenum,
1596 comment_name[incomment], linestate_name[linestate]);
1597 return (true);
1598 }
1599
1600 /*
1601 * Concatenate two strings into new memory, checking for failure.
1602 */
1603 static char *
astrcat(const char * s1,const char * s2)1604 astrcat(const char *s1, const char *s2)
1605 {
1606 char *s;
1607 int len;
1608 size_t size;
1609
1610 len = snprintf(NULL, 0, "%s%s", s1, s2);
1611 if (len < 0)
1612 err(2, "snprintf");
1613 size = (size_t)len + 1;
1614 s = (char *)malloc(size);
1615 if (s == NULL)
1616 err(2, "malloc");
1617 snprintf(s, size, "%s%s", s1, s2);
1618 return (s);
1619 }
1620
1621 /*
1622 * Duplicate a segment of a string, checking for failure.
1623 */
1624 static const char *
xstrdup(const char * start,const char * end)1625 xstrdup(const char *start, const char *end)
1626 {
1627 size_t n;
1628 char *s;
1629
1630 if (end < start) abort(); /* bug */
1631 n = (size_t)(end - start) + 1;
1632 s = malloc(n);
1633 if (s == NULL)
1634 err(2, "malloc");
1635 snprintf(s, n, "%s", start);
1636 return (s);
1637 }
1638
1639 /*
1640 * Diagnostics.
1641 */
1642 static void
debug(const char * msg,...)1643 debug(const char *msg, ...)
1644 {
1645 va_list ap;
1646
1647 if (debugging) {
1648 va_start(ap, msg);
1649 vwarnx(msg, ap);
1650 va_end(ap);
1651 }
1652 }
1653
1654 static void
error(const char * msg)1655 error(const char *msg)
1656 {
1657 if (depth == 0)
1658 warnx("%s: %d: %s", filename, linenum, msg);
1659 else
1660 warnx("%s: %d: %s (#if line %d depth %d)",
1661 filename, linenum, msg, stifline[depth], depth);
1662 closeio();
1663 errx(2, "Output may be truncated");
1664 }
1665