Annotation of src/usr.bin/unifdef/unifdef.c, Revision 1.20
1.1 deraadt 1: /*
1.16 sthen 2: * Copyright (c) 2002 - 2014 Tony Finch <dot@dotat.at>
1.1 deraadt 3: *
4: * Redistribution and use in source and binary forms, with or without
5: * modification, are permitted provided that the following conditions
6: * are met:
7: * 1. Redistributions of source code must retain the above copyright
8: * notice, this list of conditions and the following disclaimer.
9: * 2. Redistributions in binary form must reproduce the above copyright
10: * notice, this list of conditions and the following disclaimer in the
11: * documentation and/or other materials provided with the distribution.
12: *
1.16 sthen 13: * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
1.1 deraadt 14: * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
15: * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
1.16 sthen 16: * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
1.1 deraadt 17: * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
18: * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
19: * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
20: * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
21: * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
22: * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
23: * SUCH DAMAGE.
24: */
25:
26: /*
27: * unifdef - remove ifdef'ed lines
28: *
1.16 sthen 29: * This code was derived from software contributed to Berkeley by Dave Yost.
30: * It was rewritten to support ANSI C by Tony Finch. The original version
31: * of unifdef carried the 4-clause BSD copyright licence. None of its code
32: * remains in this version (though some of the names remain) so it now
33: * carries a more liberal licence.
34: *
1.1 deraadt 35: * Wishlist:
36: * provide an option which will append the name of the
37: * appropriate symbol after #else's and #endif's
38: * provide an option which will check symbols after
39: * #else's and #endif's to see that they match their
40: * corresponding #ifdef or #ifndef
1.9 deraadt 41: *
1.16 sthen 42: * These require better buffer handling, which would also make
43: * it possible to handle all "dodgy" directives correctly.
1.1 deraadt 44: */
45:
1.16 sthen 46: #include "unifdef.h"
47:
48: static const char copyright[] =
49: #include "version.h"
1.20 ! deraadt 50: "@(#) $Author: deraadt $\n"
1.16 sthen 51: "@(#) $URL: http://dotat.at/prog/unifdef $\n"
52: ;
1.1 deraadt 53:
1.7 deraadt 54: /* types of input lines: */
55: typedef enum {
1.8 deraadt 56: LT_TRUEI, /* a true #if with ignore flag */
57: LT_FALSEI, /* a false #if with ignore flag */
58: LT_IF, /* an unknown #if */
1.7 deraadt 59: LT_TRUE, /* a true #if */
60: LT_FALSE, /* a false #if */
1.8 deraadt 61: LT_ELIF, /* an unknown #elif */
1.7 deraadt 62: LT_ELTRUE, /* a true #elif */
63: LT_ELFALSE, /* a false #elif */
64: LT_ELSE, /* #else */
65: LT_ENDIF, /* #endif */
1.9 deraadt 66: LT_DODGY, /* flag: directive is not on one line */
67: LT_DODGY_LAST = LT_DODGY + LT_ENDIF,
68: LT_PLAIN, /* ordinary line */
1.8 deraadt 69: LT_EOF, /* end of file */
1.16 sthen 70: LT_ERROR, /* unevaluable #if */
1.8 deraadt 71: LT_COUNT
1.7 deraadt 72: } Linetype;
73:
1.8 deraadt 74: static char const * const linetype_name[] = {
1.9 deraadt 75: "TRUEI", "FALSEI", "IF", "TRUE", "FALSE",
76: "ELIF", "ELTRUE", "ELFALSE", "ELSE", "ENDIF",
77: "DODGY TRUEI", "DODGY FALSEI",
78: "DODGY IF", "DODGY TRUE", "DODGY FALSE",
79: "DODGY ELIF", "DODGY ELTRUE", "DODGY ELFALSE",
80: "DODGY ELSE", "DODGY ENDIF",
1.16 sthen 81: "PLAIN", "EOF", "ERROR"
1.8 deraadt 82: };
1.7 deraadt 83:
1.16 sthen 84: #define linetype_if2elif(lt) ((Linetype)(lt - LT_IF + LT_ELIF))
85: #define linetype_2dodgy(lt) ((Linetype)(lt + LT_DODGY))
86:
1.8 deraadt 87: /* state of #if processing */
1.7 deraadt 88: typedef enum {
1.8 deraadt 89: IS_OUTSIDE,
90: IS_FALSE_PREFIX, /* false #if followed by false #elifs */
91: IS_TRUE_PREFIX, /* first non-false #(el)if is true */
92: IS_PASS_MIDDLE, /* first non-false #(el)if is unknown */
93: IS_FALSE_MIDDLE, /* a false #elif after a pass state */
94: IS_TRUE_MIDDLE, /* a true #elif after a pass state */
95: IS_PASS_ELSE, /* an else after a pass state */
96: IS_FALSE_ELSE, /* an else after a true state */
97: IS_TRUE_ELSE, /* an else after only false states */
98: IS_FALSE_TRAILER, /* #elifs after a true are false */
99: IS_COUNT
100: } Ifstate;
101:
102: static char const * const ifstate_name[] = {
103: "OUTSIDE", "FALSE_PREFIX", "TRUE_PREFIX",
104: "PASS_MIDDLE", "FALSE_MIDDLE", "TRUE_MIDDLE",
105: "PASS_ELSE", "FALSE_ELSE", "TRUE_ELSE",
106: "FALSE_TRAILER"
107: };
108:
109: /* state of comment parser */
110: typedef enum {
111: NO_COMMENT = false, /* outside a comment */
112: C_COMMENT, /* in a comment like this one */
113: CXX_COMMENT, /* between // and end of line */
114: STARTING_COMMENT, /* just after slash-backslash-newline */
1.16 sthen 115: FINISHING_COMMENT, /* star-backslash-newline in a C comment */
116: CHAR_LITERAL, /* inside '' */
117: STRING_LITERAL /* inside "" */
1.7 deraadt 118: } Comment_state;
119:
1.8 deraadt 120: static char const * const comment_name[] = {
1.16 sthen 121: "NO", "C", "CXX", "STARTING", "FINISHING", "CHAR", "STRING"
1.1 deraadt 122: };
1.7 deraadt 123:
1.8 deraadt 124: /* state of preprocessor line parser */
125: typedef enum {
126: LS_START, /* only space and comments on this line */
127: LS_HASH, /* only space, comments, and a hash */
128: LS_DIRTY /* this line can't be a preprocessor line */
129: } Line_state;
1.7 deraadt 130:
1.8 deraadt 131: static char const * const linestate_name[] = {
132: "START", "HASH", "DIRTY"
133: };
1.7 deraadt 134:
135: /*
1.8 deraadt 136: * Minimum translation limits from ISO/IEC 9899:1999 5.2.4.1
1.7 deraadt 137: */
1.8 deraadt 138: #define MAXDEPTH 64 /* maximum #if nesting */
139: #define MAXLINE 4096 /* maximum length of line */
1.16 sthen 140: #define MAXSYMS 16384 /* maximum number of symbols */
1.7 deraadt 141:
142: /*
1.9 deraadt 143: * Sometimes when editing a keyword the replacement text is longer, so
144: * we leave some space at the end of the tline buffer to accommodate this.
145: */
146: #define EDITSLOP 10
147:
148: /*
1.8 deraadt 149: * Globals.
1.7 deraadt 150: */
151:
1.16 sthen 152: static bool compblank; /* -B: compress blank lines */
153: static bool lnblank; /* -b: blank deleted lines */
1.8 deraadt 154: static bool complement; /* -c: do the complement */
155: static bool debugging; /* -d: debugging reports */
1.16 sthen 156: static bool inplace; /* -m: modify in place */
1.9 deraadt 157: static bool iocccok; /* -e: fewer IOCCC errors */
1.16 sthen 158: static bool strictlogic; /* -K: keep ambiguous #ifs */
1.8 deraadt 159: static bool killconsts; /* -k: eval constant #ifs */
1.16 sthen 160: static bool lnnum; /* -n: add #line directives */
1.8 deraadt 161: static bool symlist; /* -s: output symbol list */
1.16 sthen 162: static bool symdepth; /* -S: output symbol depth */
1.8 deraadt 163: static bool text; /* -t: this is a text file */
164:
165: static const char *symname[MAXSYMS]; /* symbol name */
166: static const char *value[MAXSYMS]; /* -Dsym=value */
167: static bool ignore[MAXSYMS]; /* -iDsym or -iUsym */
168: static int nsyms; /* number of symbols */
169:
170: static FILE *input; /* input file pointer */
171: static const char *filename; /* input file name */
172: static int linenum; /* current line number */
1.16 sthen 173: static const char *linefile; /* file name for #line */
174: static FILE *output; /* output file pointer */
175: static const char *ofilename; /* output file name */
176: static const char *backext; /* backup extension */
177: static char *tempname; /* avoid splatting input */
1.8 deraadt 178:
1.9 deraadt 179: static char tline[MAXLINE+EDITSLOP];/* input buffer plus space */
1.8 deraadt 180: static char *keyword; /* used for editing #elif's */
181:
1.16 sthen 182: /*
183: * When processing a file, the output's newline style will match the
184: * input's, and unifdef correctly handles CRLF or LF endings whatever
185: * the platform's native style. The stdio streams are opened in binary
186: * mode to accommodate platforms whose native newline style is CRLF.
187: * When the output isn't a processed input file (when it is error /
188: * debug / diagnostic messages) then unifdef uses native line endings.
189: */
190:
191: static const char *newline; /* input file format */
192: static const char newline_unix[] = "\n";
193: static const char newline_crlf[] = "\r\n";
194:
1.8 deraadt 195: static Comment_state incomment; /* comment parser state */
196: static Line_state linestate; /* #if line parser state */
197: static Ifstate ifstate[MAXDEPTH]; /* #if processor state */
198: static bool ignoring[MAXDEPTH]; /* ignore comments state */
199: static int stifline[MAXDEPTH]; /* start of current #if */
200: static int depth; /* current #if nesting */
1.16 sthen 201: static int delcount; /* count of deleted lines */
202: static unsigned blankcount; /* count of blank lines */
203: static unsigned blankmax; /* maximum recent blankcount */
204: static bool constexpr; /* constant #if expression */
205: static bool zerosyms; /* to format symdepth output */
206: static bool firstsym; /* ditto */
1.8 deraadt 207:
1.16 sthen 208: static int exitmode; /* exit status mode */
1.8 deraadt 209: static int exitstat; /* program exit status */
210:
1.16 sthen 211: static void addsym1(bool, bool, char *);
212: static void addsym2(bool, const char *, const char *);
213: static char *astrcat(const char *, const char *);
214: static void cleantemp(void);
215: static void closeio(void);
1.8 deraadt 216: static void debug(const char *, ...);
1.16 sthen 217: static void debugsym(const char *, int);
218: static bool defundef(void);
219: static void defundefile(const char *);
220: static void done(void);
1.8 deraadt 221: static void error(const char *);
1.16 sthen 222: static int findsym(const char **);
1.8 deraadt 223: static void flushline(bool);
1.16 sthen 224: static void hashline(void);
225: static void help(void);
1.8 deraadt 226: static Linetype ifeval(const char **);
1.9 deraadt 227: static void ignoreoff(void);
228: static void ignoreon(void);
1.16 sthen 229: static void indirectsym(void);
1.9 deraadt 230: static void keywordedit(const char *);
1.16 sthen 231: static const char *matchsym(const char *, const char *);
1.8 deraadt 232: static void nest(void);
1.16 sthen 233: static Linetype parseline(void);
1.8 deraadt 234: static void process(void);
1.16 sthen 235: static void processinout(const char *, const char *);
236: static const char *skipargs(const char *);
1.8 deraadt 237: static const char *skipcomment(const char *);
1.16 sthen 238: static const char *skiphash(void);
239: static const char *skipline(const char *);
1.8 deraadt 240: static const char *skipsym(const char *);
241: static void state(Ifstate);
1.16 sthen 242: static void unnest(void);
1.8 deraadt 243: static void usage(void);
1.16 sthen 244: static void version(void);
245: static const char *xstrdup(const char *, const char *);
1.7 deraadt 246:
1.16 sthen 247: #define endsym(c) (!isalnum((unsigned char)c) && c != '_')
1.7 deraadt 248:
1.8 deraadt 249: /*
250: * The main program.
251: */
1.7 deraadt 252: int
253: main(int argc, char *argv[])
254: {
1.19 deraadt 255: const char *errstr;
1.7 deraadt 256: int opt;
257:
1.16 sthen 258: while ((opt = getopt(argc, argv, "i:D:U:f:I:M:o:x:bBcdehKklmnsStV")) != -1)
1.7 deraadt 259: switch (opt) {
260: case 'i': /* treat stuff controlled by these symbols as text */
261: /*
262: * For strict backwards-compatibility the U or D
263: * should be immediately after the -i but it doesn't
264: * matter much if we relax that requirement.
265: */
266: opt = *optarg++;
267: if (opt == 'D')
1.16 sthen 268: addsym1(true, true, optarg);
1.7 deraadt 269: else if (opt == 'U')
1.16 sthen 270: addsym1(true, false, optarg);
1.7 deraadt 271: else
272: usage();
273: break;
274: case 'D': /* define a symbol */
1.16 sthen 275: addsym1(false, true, optarg);
1.7 deraadt 276: break;
277: case 'U': /* undef a symbol */
1.16 sthen 278: addsym1(false, false, optarg);
279: break;
280: case 'I': /* no-op for compatibility with cpp */
281: break;
282: case 'b': /* blank deleted lines instead of omitting them */
283: case 'l': /* backwards compatibility */
284: lnblank = true;
1.7 deraadt 285: break;
1.16 sthen 286: case 'B': /* compress blank lines around removed section */
287: compblank = true;
1.11 avsm 288: break;
1.7 deraadt 289: case 'c': /* treat -D as -U and vice versa */
290: complement = true;
291: break;
1.8 deraadt 292: case 'd':
293: debugging = true;
294: break;
1.9 deraadt 295: case 'e': /* fewer errors from dodgy lines */
296: iocccok = true;
297: break;
1.16 sthen 298: case 'f': /* definitions file */
299: defundefile(optarg);
300: break;
301: case 'h':
302: help();
303: break;
304: case 'K': /* keep ambiguous #ifs */
305: strictlogic = true;
306: break;
1.7 deraadt 307: case 'k': /* process constant #ifs */
308: killconsts = true;
309: break;
1.16 sthen 310: case 'm': /* modify in place */
311: inplace = true;
312: break;
313: case 'M': /* modify in place and keep backup */
314: inplace = true;
315: backext = optarg;
316: break;
317: case 'n': /* add #line directive after deleted lines */
318: lnnum = true;
319: break;
320: case 'o': /* output to a file */
321: ofilename = optarg;
1.7 deraadt 322: break;
323: case 's': /* only output list of symbols that control #ifs */
324: symlist = true;
325: break;
1.16 sthen 326: case 'S': /* list symbols with their nesting depth */
327: symlist = symdepth = true;
328: break;
1.8 deraadt 329: case 't': /* don't parse C comments */
1.7 deraadt 330: text = true;
331: break;
1.16 sthen 332: case 'V':
333: version();
334: break;
335: case 'x':
1.19 deraadt 336: exitmode = strtonum(optarg, 0, 2, &errstr);
337: if (errstr)
338: errx(1, "-x %s: %s", optarg, errstr);
1.16 sthen 339: break;
1.7 deraadt 340: default:
341: usage();
342: }
343: argc -= optind;
344: argv += optind;
1.16 sthen 345: if (compblank && lnblank)
346: errx(2, "-B and -b are mutually exclusive");
347: if (symlist && (ofilename != NULL || inplace || argc > 1))
348: errx(2, "-s only works with one input file");
349: if (argc > 1 && ofilename != NULL)
350: errx(2, "-o cannot be used with multiple input files");
351: if (argc > 1 && !inplace)
352: errx(2, "multiple input files require -m or -M");
353: if (argc == 0)
354: argc = 1;
355: if (argc == 1 && !inplace && ofilename == NULL)
356: ofilename = "-";
357: indirectsym();
358:
359: atexit(cleantemp);
360: if (ofilename != NULL)
361: processinout(*argv, ofilename);
362: else while (argc-- > 0) {
363: processinout(*argv, *argv);
364: argv++;
365: }
366: switch(exitmode) {
367: case(0): exit(exitstat);
368: case(1): exit(!exitstat);
369: case(2): exit(0);
370: default: abort(); /* bug */
1.7 deraadt 371: }
1.16 sthen 372: }
373:
374: /*
375: * File logistics.
376: */
377: static void
378: processinout(const char *ifn, const char *ofn)
379: {
380: struct stat st;
381:
382: if (ifn == NULL || strcmp(ifn, "-") == 0) {
383: filename = "[stdin]";
384: linefile = NULL;
385: input = fbinmode(stdin);
1.7 deraadt 386: } else {
1.16 sthen 387: filename = ifn;
388: linefile = ifn;
389: input = fopen(ifn, "rb");
390: if (input == NULL)
391: err(2, "can't open %s", ifn);
392: }
393: if (strcmp(ofn, "-") == 0) {
394: output = fbinmode(stdout);
395: process();
396: return;
397: }
398: if (stat(ofn, &st) < 0) {
399: output = fopen(ofn, "wb");
400: if (output == NULL)
401: err(2, "can't create %s", ofn);
1.8 deraadt 402: process();
1.16 sthen 403: return;
404: }
405:
406: tempname = astrcat(ofn, ".XXXXXX");
407: output = mktempmode(tempname, st.st_mode);
408: if (output == NULL)
409: err(2, "can't create %s", tempname);
410:
411: process();
412:
413: if (backext != NULL) {
414: char *backname = astrcat(ofn, backext);
415: if (rename(ofn, backname) < 0)
416: err(2, "can't rename \"%s\" to \"%s\"", ofn, backname);
417: free(backname);
418: }
419: if (replace(tempname, ofn) < 0)
420: err(2, "can't rename \"%s\" to \"%s\"", tempname, ofn);
421: free(tempname);
422: tempname = NULL;
423: }
424:
425: /*
426: * For cleaning up if there is an error.
427: */
428: static void
429: cleantemp(void)
430: {
431: if (tempname != NULL)
432: remove(tempname);
433: }
434:
435: /*
436: * Self-identification functions.
437: */
438:
439: static void
440: version(void)
441: {
442: const char *c = copyright;
443: for (;;) {
444: while (*++c != '$')
445: if (*c == '\0')
446: exit(0);
447: while (*++c != '$')
448: putc(*c, stderr);
449: putc('\n', stderr);
1.7 deraadt 450: }
1.16 sthen 451: }
1.7 deraadt 452:
1.16 sthen 453: static void
454: synopsis(FILE *fp)
455: {
456: fprintf(fp,
1.17 jmc 457: "usage: unifdef [-BbcdehKkmnSstV] [-[i]Dsym[=val]] [-[i]Usym] [-f defile]\n"
458: " [-M backext] [-o outfile] [-x 0 | 1 | 2] file ...\n");
1.7 deraadt 459: }
1.1 deraadt 460:
1.8 deraadt 461: static void
1.7 deraadt 462: usage(void)
1.1 deraadt 463: {
1.16 sthen 464: synopsis(stderr);
1.8 deraadt 465: exit(2);
466: }
467:
1.16 sthen 468: static void
469: help(void)
470: {
471: synopsis(stdout);
472: printf(
473: " -Dsym=val define preprocessor symbol with given value\n"
474: " -Dsym define preprocessor symbol with value 1\n"
475: " -Usym preprocessor symbol is undefined\n"
476: " -iDsym=val \\ ignore C strings and comments\n"
477: " -iDsym ) in sections controlled by these\n"
478: " -iUsym / preprocessor symbols\n"
479: " -fpath file containing #define and #undef directives\n"
480: " -b blank lines instead of deleting them\n"
481: " -B compress blank lines around deleted section\n"
482: " -c complement (invert) keep vs. delete\n"
483: " -d debugging mode\n"
484: " -e ignore multiline preprocessor directives\n"
485: " -h print help\n"
486: " -Ipath extra include file path (ignored)\n"
487: " -K disable && and || short-circuiting\n"
488: " -k process constant #if expressions\n"
489: " -Mext modify in place and keep backups\n"
490: " -m modify input files in place\n"
491: " -n add #line directives to output\n"
492: " -opath output file name\n"
493: " -S list #if control symbols with nesting\n"
494: " -s list #if control symbols\n"
495: " -t ignore C strings and comments\n"
496: " -V print version\n"
497: " -x{012} exit status mode\n"
498: );
499: exit(0);
500: }
501:
1.8 deraadt 502: /*
503: * A state transition function alters the global #if processing state
504: * in a particular way. The table below is indexed by the current
1.16 sthen 505: * processing state and the type of the current line.
1.8 deraadt 506: *
507: * Nesting is handled by keeping a stack of states; some transition
1.9 deraadt 508: * functions increase or decrease the depth. They also maintain the
1.8 deraadt 509: * ignore state on a stack. In some complicated cases they have to
510: * alter the preprocessor directive, as follows.
511: *
512: * When we have processed a group that starts off with a known-false
513: * #if/#elif sequence (which has therefore been deleted) followed by a
1.9 deraadt 514: * #elif that we don't understand and therefore must keep, we edit the
1.16 sthen 515: * latter into a #if to keep the nesting correct. We use memcpy() to
516: * overwrite the 4 byte token "elif" with "if " without a '\0' byte.
1.8 deraadt 517: *
518: * When we find a true #elif in a group, the following block will
519: * always be kept and the rest of the sequence after the next #elif or
1.9 deraadt 520: * #else will be discarded. We edit the #elif into a #else and the
1.8 deraadt 521: * following directive to #endif since this has the desired behaviour.
1.9 deraadt 522: *
523: * "Dodgy" directives are split across multiple lines, the most common
524: * example being a multi-line comment hanging off the right of the
525: * directive. We can handle them correctly only if there is no change
526: * from printing to dropping (or vice versa) caused by that directive.
527: * If the directive is the first of a group we have a choice between
528: * failing with an error, or passing it through unchanged instead of
529: * evaluating it. The latter is not the default to avoid questions from
530: * users about unifdef unexpectedly leaving behind preprocessor directives.
1.8 deraadt 531: */
532: typedef void state_fn(void);
533:
534: /* report an error */
1.16 sthen 535: static void Eelif (void) { error("Inappropriate #elif"); }
536: static void Eelse (void) { error("Inappropriate #else"); }
537: static void Eendif(void) { error("Inappropriate #endif"); }
538: static void Eeof (void) { error("Premature EOF"); }
539: static void Eioccc(void) { error("Obfuscated preprocessor control line"); }
1.8 deraadt 540: /* plain line handling */
1.16 sthen 541: static void print (void) { flushline(true); }
542: static void drop (void) { flushline(false); }
1.8 deraadt 543: /* output lacks group's start line */
1.16 sthen 544: static void Strue (void) { drop(); ignoreoff(); state(IS_TRUE_PREFIX); }
545: static void Sfalse(void) { drop(); ignoreoff(); state(IS_FALSE_PREFIX); }
546: static void Selse (void) { drop(); state(IS_TRUE_ELSE); }
1.8 deraadt 547: /* print/pass this block */
1.16 sthen 548: static void Pelif (void) { print(); ignoreoff(); state(IS_PASS_MIDDLE); }
549: static void Pelse (void) { print(); state(IS_PASS_ELSE); }
550: static void Pendif(void) { print(); unnest(); }
1.8 deraadt 551: /* discard this block */
1.16 sthen 552: static void Dfalse(void) { drop(); ignoreoff(); state(IS_FALSE_TRAILER); }
553: static void Delif (void) { drop(); ignoreoff(); state(IS_FALSE_MIDDLE); }
554: static void Delse (void) { drop(); state(IS_FALSE_ELSE); }
555: static void Dendif(void) { drop(); unnest(); }
1.8 deraadt 556: /* first line of group */
1.16 sthen 557: static void Fdrop (void) { nest(); Dfalse(); }
558: static void Fpass (void) { nest(); Pelif(); }
559: static void Ftrue (void) { nest(); Strue(); }
560: static void Ffalse(void) { nest(); Sfalse(); }
1.9 deraadt 561: /* variable pedantry for obfuscated lines */
1.16 sthen 562: static void Oiffy (void) { if (!iocccok) Eioccc(); Fpass(); ignoreon(); }
563: static void Oif (void) { if (!iocccok) Eioccc(); Fpass(); }
564: static void Oelif (void) { if (!iocccok) Eioccc(); Pelif(); }
1.8 deraadt 565: /* ignore comments in this block */
1.16 sthen 566: static void Idrop (void) { Fdrop(); ignoreon(); }
567: static void Itrue (void) { Ftrue(); ignoreon(); }
568: static void Ifalse(void) { Ffalse(); ignoreon(); }
569: /* modify this line */
570: static void Mpass (void) { memcpy(keyword, "if ", 4); Pelif(); }
571: static void Mtrue (void) { keywordedit("else"); state(IS_TRUE_MIDDLE); }
572: static void Melif (void) { keywordedit("endif"); state(IS_FALSE_TRAILER); }
573: static void Melse (void) { keywordedit("endif"); state(IS_FALSE_ELSE); }
1.8 deraadt 574:
575: static state_fn * const trans_table[IS_COUNT][LT_COUNT] = {
576: /* IS_OUTSIDE */
1.9 deraadt 577: { Itrue, Ifalse,Fpass, Ftrue, Ffalse,Eelif, Eelif, Eelif, Eelse, Eendif,
578: Oiffy, Oiffy, Fpass, Oif, Oif, Eelif, Eelif, Eelif, Eelse, Eendif,
1.16 sthen 579: print, done, abort },
1.8 deraadt 580: /* IS_FALSE_PREFIX */
1.9 deraadt 581: { Idrop, Idrop, Fdrop, Fdrop, Fdrop, Mpass, Strue, Sfalse,Selse, Dendif,
582: Idrop, Idrop, Fdrop, Fdrop, Fdrop, Mpass, Eioccc,Eioccc,Eioccc,Eioccc,
1.16 sthen 583: drop, Eeof, abort },
1.8 deraadt 584: /* IS_TRUE_PREFIX */
1.9 deraadt 585: { Itrue, Ifalse,Fpass, Ftrue, Ffalse,Dfalse,Dfalse,Dfalse,Delse, Dendif,
586: Oiffy, Oiffy, Fpass, Oif, Oif, Eioccc,Eioccc,Eioccc,Eioccc,Eioccc,
1.16 sthen 587: print, Eeof, abort },
1.8 deraadt 588: /* IS_PASS_MIDDLE */
1.9 deraadt 589: { Itrue, Ifalse,Fpass, Ftrue, Ffalse,Pelif, Mtrue, Delif, Pelse, Pendif,
590: Oiffy, Oiffy, Fpass, Oif, Oif, Pelif, Oelif, Oelif, Pelse, Pendif,
1.16 sthen 591: print, Eeof, abort },
1.8 deraadt 592: /* IS_FALSE_MIDDLE */
1.9 deraadt 593: { Idrop, Idrop, Fdrop, Fdrop, Fdrop, Pelif, Mtrue, Delif, Pelse, Pendif,
594: Idrop, Idrop, Fdrop, Fdrop, Fdrop, Eioccc,Eioccc,Eioccc,Eioccc,Eioccc,
1.16 sthen 595: drop, Eeof, abort },
1.8 deraadt 596: /* IS_TRUE_MIDDLE */
1.9 deraadt 597: { Itrue, Ifalse,Fpass, Ftrue, Ffalse,Melif, Melif, Melif, Melse, Pendif,
598: Oiffy, Oiffy, Fpass, Oif, Oif, Eioccc,Eioccc,Eioccc,Eioccc,Pendif,
1.16 sthen 599: print, Eeof, abort },
1.8 deraadt 600: /* IS_PASS_ELSE */
1.9 deraadt 601: { Itrue, Ifalse,Fpass, Ftrue, Ffalse,Eelif, Eelif, Eelif, Eelse, Pendif,
602: Oiffy, Oiffy, Fpass, Oif, Oif, Eelif, Eelif, Eelif, Eelse, Pendif,
1.16 sthen 603: print, Eeof, abort },
1.8 deraadt 604: /* IS_FALSE_ELSE */
1.9 deraadt 605: { Idrop, Idrop, Fdrop, Fdrop, Fdrop, Eelif, Eelif, Eelif, Eelse, Dendif,
606: Idrop, Idrop, Fdrop, Fdrop, Fdrop, Eelif, Eelif, Eelif, Eelse, Eioccc,
1.16 sthen 607: drop, Eeof, abort },
1.8 deraadt 608: /* IS_TRUE_ELSE */
1.9 deraadt 609: { Itrue, Ifalse,Fpass, Ftrue, Ffalse,Eelif, Eelif, Eelif, Eelse, Dendif,
610: Oiffy, Oiffy, Fpass, Oif, Oif, Eelif, Eelif, Eelif, Eelse, Eioccc,
1.16 sthen 611: print, Eeof, abort },
1.8 deraadt 612: /* IS_FALSE_TRAILER */
1.9 deraadt 613: { Idrop, Idrop, Fdrop, Fdrop, Fdrop, Dfalse,Dfalse,Dfalse,Delse, Dendif,
614: Idrop, Idrop, Fdrop, Fdrop, Fdrop, Dfalse,Dfalse,Dfalse,Delse, Eioccc,
1.16 sthen 615: drop, Eeof, abort }
1.9 deraadt 616: /*TRUEI FALSEI IF TRUE FALSE ELIF ELTRUE ELFALSE ELSE ENDIF
617: TRUEI FALSEI IF TRUE FALSE ELIF ELTRUE ELFALSE ELSE ENDIF (DODGY)
1.16 sthen 618: PLAIN EOF ERROR */
1.8 deraadt 619: };
620:
621: /*
622: * State machine utility functions
623: */
624: static void
1.9 deraadt 625: ignoreoff(void)
626: {
1.16 sthen 627: if (depth == 0)
628: abort(); /* bug */
1.9 deraadt 629: ignoring[depth] = ignoring[depth-1];
630: }
631: static void
632: ignoreon(void)
633: {
634: ignoring[depth] = true;
635: }
636: static void
637: keywordedit(const char *replacement)
638: {
1.16 sthen 639: snprintf(keyword, tline + sizeof(tline) - keyword,
640: "%s%s", replacement, newline);
1.9 deraadt 641: print();
642: }
643: static void
1.8 deraadt 644: nest(void)
645: {
1.16 sthen 646: if (depth > MAXDEPTH-1)
647: abort(); /* bug */
648: if (depth == MAXDEPTH-1)
649: error("Too many levels of nesting");
1.8 deraadt 650: depth += 1;
651: stifline[depth] = linenum;
652: }
1.16 sthen 653: static void
654: unnest(void)
655: {
656: if (depth == 0)
657: abort(); /* bug */
658: depth -= 1;
659: }
1.8 deraadt 660: static void
661: state(Ifstate is)
662: {
663: ifstate[depth] = is;
664: }
665:
1.7 deraadt 666: /*
1.16 sthen 667: * The last state transition function. When this is called,
668: * lineval == LT_EOF, so the process() loop will terminate.
669: */
670: static void
671: done(void)
672: {
673: if (incomment)
674: error("EOF in comment");
675: closeio();
676: }
677:
678: /*
1.8 deraadt 679: * Write a line to the output or not, according to command line options.
1.16 sthen 680: * If writing fails, closeio() will print the error and exit.
1.8 deraadt 681: */
682: static void
683: flushline(bool keep)
684: {
685: if (symlist)
686: return;
1.16 sthen 687: if (keep ^ complement) {
688: bool blankline = tline[strspn(tline, " \t\r\n")] == '\0';
689: if (blankline && compblank && blankcount != blankmax) {
690: delcount += 1;
691: blankcount += 1;
692: } else {
693: if (lnnum && delcount > 0)
694: hashline();
695: if (fputs(tline, output) == EOF)
696: closeio();
697: delcount = 0;
698: blankmax = blankcount = blankline ? blankcount + 1 : 0;
699: }
700: } else {
701: if (lnblank && fputs(newline, output) == EOF)
702: closeio();
1.8 deraadt 703: exitstat = 1;
1.16 sthen 704: delcount += 1;
705: blankcount = 0;
1.7 deraadt 706: }
1.16 sthen 707: if (debugging && fflush(output) == EOF)
708: closeio();
709: }
710:
711: /*
712: * Format of #line directives depends on whether we know the input filename.
713: */
714: static void
715: hashline(void)
716: {
717: int e;
718:
719: if (linefile == NULL)
720: e = fprintf(output, "#line %d%s", linenum, newline);
721: else
722: e = fprintf(output, "#line %d \"%s\"%s",
723: linenum, linefile, newline);
724: if (e < 0)
725: closeio();
726: }
727:
728: /*
729: * Flush the output and handle errors.
730: */
731: static void
732: closeio(void)
733: {
734: /* Tidy up after findsym(). */
735: if (symdepth && !zerosyms)
736: printf("\n");
737: if (output != NULL && (ferror(output) || fclose(output) == EOF))
738: err(2, "%s: can't write to output", filename);
739: fclose(input);
1.7 deraadt 740: }
1.3 deraadt 741:
1.7 deraadt 742: /*
1.8 deraadt 743: * The driver for the state machine.
1.7 deraadt 744: */
1.8 deraadt 745: static void
746: process(void)
1.7 deraadt 747: {
1.16 sthen 748: Linetype lineval = LT_PLAIN;
749: /* When compressing blank lines, act as if the file
750: is preceded by a large number of blank lines. */
751: blankmax = blankcount = 1000;
752: zerosyms = true;
753: newline = NULL;
754: linenum = 0;
755: while (lineval != LT_EOF) {
756: lineval = parseline();
757: trans_table[ifstate[depth]][lineval]();
758: debug("process line %d %s -> %s depth %d",
759: linenum, linetype_name[lineval],
1.8 deraadt 760: ifstate_name[ifstate[depth]], depth);
1.1 deraadt 761: }
762: }
763:
1.7 deraadt 764: /*
1.8 deraadt 765: * Parse a line and determine its type. We keep the preprocessor line
1.16 sthen 766: * parser state between calls in the global variable linestate, with
767: * help from skipcomment().
1.7 deraadt 768: */
1.8 deraadt 769: static Linetype
1.16 sthen 770: parseline(void)
1.3 deraadt 771: {
1.7 deraadt 772: const char *cp;
1.8 deraadt 773: int cursym;
1.3 deraadt 774: Linetype retval;
1.8 deraadt 775: Comment_state wascomment;
1.3 deraadt 776:
1.16 sthen 777: wascomment = incomment;
778: cp = skiphash();
779: if (cp == NULL)
1.8 deraadt 780: return (LT_EOF);
1.16 sthen 781: if (newline == NULL) {
782: if (strrchr(tline, '\n') == strrchr(tline, '\r') + 1)
783: newline = newline_crlf;
784: else
785: newline = newline_unix;
786: }
787: if (*cp == '\0') {
788: retval = LT_PLAIN;
789: goto done;
1.8 deraadt 790: }
1.16 sthen 791: keyword = tline + (cp - tline);
792: if ((cp = matchsym("ifdef", keyword)) != NULL ||
793: (cp = matchsym("ifndef", keyword)) != NULL) {
794: cp = skipcomment(cp);
795: if ((cursym = findsym(&cp)) < 0)
796: retval = LT_IF;
797: else {
798: retval = (keyword[2] == 'n')
799: ? LT_FALSE : LT_TRUE;
800: if (value[cursym] == NULL)
801: retval = (retval == LT_TRUE)
802: ? LT_FALSE : LT_TRUE;
803: if (ignore[cursym])
804: retval = (retval == LT_TRUE)
805: ? LT_TRUEI : LT_FALSEI;
806: }
807: } else if ((cp = matchsym("if", keyword)) != NULL)
808: retval = ifeval(&cp);
809: else if ((cp = matchsym("elif", keyword)) != NULL)
810: retval = linetype_if2elif(ifeval(&cp));
811: else if ((cp = matchsym("else", keyword)) != NULL)
812: retval = LT_ELSE;
813: else if ((cp = matchsym("endif", keyword)) != NULL)
814: retval = LT_ENDIF;
815: else {
816: cp = skipsym(keyword);
1.9 deraadt 817: /* no way can we deal with a continuation inside a keyword */
1.16 sthen 818: if (strncmp(cp, "\\\r\n", 3) == 0 ||
819: strncmp(cp, "\\\n", 2) == 0)
1.8 deraadt 820: Eioccc();
1.16 sthen 821: cp = skipline(cp);
822: retval = LT_PLAIN;
823: goto done;
824: }
825: cp = skipcomment(cp);
826: if (*cp != '\0') {
827: cp = skipline(cp);
828: if (retval == LT_TRUE || retval == LT_FALSE ||
829: retval == LT_TRUEI || retval == LT_FALSEI)
830: retval = LT_IF;
831: if (retval == LT_ELTRUE || retval == LT_ELFALSE)
832: retval = LT_ELIF;
833: }
834: /* the following can happen if the last line of the file lacks a
835: newline or if there is too much whitespace in a directive */
836: if (linestate == LS_HASH) {
837: long len = cp - tline;
838: if (fgets(tline + len, MAXLINE - len, input) == NULL) {
839: if (ferror(input))
840: err(2, "can't read %s", filename);
841: /* append the missing newline at eof */
1.18 miod 842: strlcpy(tline + len, newline, sizeof(tline) - len);
1.16 sthen 843: cp += strlen(newline);
844: linestate = LS_START;
845: } else {
1.8 deraadt 846: linestate = LS_DIRTY;
1.7 deraadt 847: }
848: }
1.16 sthen 849: if (retval != LT_PLAIN && (wascomment || linestate != LS_START)) {
850: retval = linetype_2dodgy(retval);
851: linestate = LS_DIRTY;
1.8 deraadt 852: }
1.16 sthen 853: done:
854: debug("parser line %d state %s comment %s line", linenum,
1.8 deraadt 855: comment_name[incomment], linestate_name[linestate]);
1.7 deraadt 856: return (retval);
857: }
858:
859: /*
1.16 sthen 860: * These are the binary operators that are supported by the expression
861: * evaluator.
1.7 deraadt 862: */
1.16 sthen 863: static Linetype op_strict(long *p, long v, Linetype at, Linetype bt) {
864: if(at == LT_IF || bt == LT_IF) return (LT_IF);
865: return (*p = v, v ? LT_TRUE : LT_FALSE);
866: }
867: static Linetype op_lt(long *p, Linetype at, long a, Linetype bt, long b) {
868: return op_strict(p, a < b, at, bt);
1.8 deraadt 869: }
1.16 sthen 870: static Linetype op_gt(long *p, Linetype at, long a, Linetype bt, long b) {
871: return op_strict(p, a > b, at, bt);
1.8 deraadt 872: }
1.16 sthen 873: static Linetype op_le(long *p, Linetype at, long a, Linetype bt, long b) {
874: return op_strict(p, a <= b, at, bt);
1.8 deraadt 875: }
1.16 sthen 876: static Linetype op_ge(long *p, Linetype at, long a, Linetype bt, long b) {
877: return op_strict(p, a >= b, at, bt);
1.8 deraadt 878: }
1.16 sthen 879: static Linetype op_eq(long *p, Linetype at, long a, Linetype bt, long b) {
880: return op_strict(p, a == b, at, bt);
1.8 deraadt 881: }
1.16 sthen 882: static Linetype op_ne(long *p, Linetype at, long a, Linetype bt, long b) {
883: return op_strict(p, a != b, at, bt);
1.8 deraadt 884: }
1.16 sthen 885: static Linetype op_or(long *p, Linetype at, long a, Linetype bt, long b) {
886: if (!strictlogic && (at == LT_TRUE || bt == LT_TRUE))
887: return (*p = 1, LT_TRUE);
888: return op_strict(p, a || b, at, bt);
1.8 deraadt 889: }
1.16 sthen 890: static Linetype op_and(long *p, Linetype at, long a, Linetype bt, long b) {
891: if (!strictlogic && (at == LT_FALSE || bt == LT_FALSE))
892: return (*p = 0, LT_FALSE);
893: return op_strict(p, a && b, at, bt);
1.7 deraadt 894: }
895:
896: /*
1.8 deraadt 897: * An evaluation function takes three arguments, as follows: (1) a pointer to
898: * an element of the precedence table which lists the operators at the current
899: * level of precedence; (2) a pointer to an integer which will receive the
900: * value of the expression; and (3) a pointer to a char* that points to the
901: * expression to be evaluated and that is updated to the end of the expression
902: * when evaluation is complete. The function returns LT_FALSE if the value of
1.16 sthen 903: * the expression is zero, LT_TRUE if it is non-zero, LT_IF if the expression
904: * depends on an unknown symbol, or LT_ERROR if there is a parse failure.
1.7 deraadt 905: */
1.8 deraadt 906: struct ops;
907:
1.16 sthen 908: typedef Linetype eval_fn(const struct ops *, long *, const char **);
1.8 deraadt 909:
910: static eval_fn eval_table, eval_unary;
911:
912: /*
913: * The precedence table. Expressions involving binary operators are evaluated
914: * in a table-driven way by eval_table. When it evaluates a subexpression it
915: * calls the inner function with its first argument pointing to the next
916: * element of the table. Innermost expressions have special non-table-driven
917: * handling.
918: */
1.16 sthen 919: struct op {
920: const char *str;
921: Linetype (*fn)(long *, Linetype, long, Linetype, long);
922: };
923: struct ops {
1.8 deraadt 924: eval_fn *inner;
1.16 sthen 925: struct op op[5];
926: };
927: static const struct ops eval_ops[] = {
1.8 deraadt 928: { eval_table, { { "||", op_or } } },
929: { eval_table, { { "&&", op_and } } },
930: { eval_table, { { "==", op_eq },
931: { "!=", op_ne } } },
932: { eval_unary, { { "<=", op_le },
933: { ">=", op_ge },
934: { "<", op_lt },
935: { ">", op_gt } } }
936: };
1.7 deraadt 937:
1.16 sthen 938: /* Current operator precedence level */
939: static long prec(const struct ops *ops)
940: {
941: return (ops - eval_ops);
942: }
943:
1.7 deraadt 944: /*
945: * Function for evaluating the innermost parts of expressions,
1.16 sthen 946: * viz. !expr (expr) number defined(symbol) symbol
947: * We reset the constexpr flag in the last two cases.
1.7 deraadt 948: */
1.8 deraadt 949: static Linetype
1.16 sthen 950: eval_unary(const struct ops *ops, long *valp, const char **cpp)
1.7 deraadt 951: {
952: const char *cp;
953: char *ep;
954: int sym;
1.16 sthen 955: bool defparen;
956: Linetype lt;
1.7 deraadt 957:
958: cp = skipcomment(*cpp);
1.8 deraadt 959: if (*cp == '!') {
1.16 sthen 960: debug("eval%d !", prec(ops));
1.7 deraadt 961: cp++;
1.16 sthen 962: lt = eval_unary(ops, valp, &cp);
963: if (lt == LT_ERROR)
964: return (LT_ERROR);
965: if (lt != LT_IF) {
966: *valp = !*valp;
967: lt = *valp ? LT_TRUE : LT_FALSE;
968: }
1.7 deraadt 969: } else if (*cp == '(') {
970: cp++;
1.16 sthen 971: debug("eval%d (", prec(ops));
972: lt = eval_table(eval_ops, valp, &cp);
973: if (lt == LT_ERROR)
974: return (LT_ERROR);
1.7 deraadt 975: cp = skipcomment(cp);
976: if (*cp++ != ')')
1.16 sthen 977: return (LT_ERROR);
1.7 deraadt 978: } else if (isdigit((unsigned char)*cp)) {
1.16 sthen 979: debug("eval%d number", prec(ops));
1.7 deraadt 980: *valp = strtol(cp, &ep, 0);
1.16 sthen 981: if (ep == cp)
982: return (LT_ERROR);
983: lt = *valp ? LT_TRUE : LT_FALSE;
984: cp = ep;
985: } else if (matchsym("defined", cp) != NULL) {
1.7 deraadt 986: cp = skipcomment(cp+7);
1.16 sthen 987: if (*cp == '(') {
988: cp = skipcomment(cp+1);
989: defparen = true;
990: } else {
991: defparen = false;
992: }
993: sym = findsym(&cp);
1.7 deraadt 994: cp = skipcomment(cp);
1.16 sthen 995: if (defparen && *cp++ != ')') {
996: debug("eval%d defined missing ')'", prec(ops));
997: return (LT_ERROR);
998: }
999: if (sym < 0) {
1000: debug("eval%d defined unknown", prec(ops));
1001: lt = LT_IF;
1002: } else {
1003: debug("eval%d defined %s", prec(ops), symname[sym]);
1004: *valp = (value[sym] != NULL);
1005: lt = *valp ? LT_TRUE : LT_FALSE;
1006: }
1007: constexpr = false;
1.7 deraadt 1008: } else if (!endsym(*cp)) {
1.16 sthen 1009: debug("eval%d symbol", prec(ops));
1010: sym = findsym(&cp);
1011: if (sym < 0) {
1012: lt = LT_IF;
1013: cp = skipargs(cp);
1014: } else if (value[sym] == NULL) {
1.7 deraadt 1015: *valp = 0;
1.16 sthen 1016: lt = LT_FALSE;
1017: } else {
1.7 deraadt 1018: *valp = strtol(value[sym], &ep, 0);
1019: if (*ep != '\0' || ep == value[sym])
1.16 sthen 1020: return (LT_ERROR);
1021: lt = *valp ? LT_TRUE : LT_FALSE;
1022: cp = skipargs(cp);
1.7 deraadt 1023: }
1.16 sthen 1024: constexpr = false;
1025: } else {
1026: debug("eval%d bad expr", prec(ops));
1027: return (LT_ERROR);
1028: }
1.7 deraadt 1029:
1030: *cpp = cp;
1.16 sthen 1031: debug("eval%d = %d", prec(ops), *valp);
1032: return (lt);
1.7 deraadt 1033: }
1034:
1035: /*
1036: * Table-driven evaluation of binary operators.
1037: */
1.8 deraadt 1038: static Linetype
1.16 sthen 1039: eval_table(const struct ops *ops, long *valp, const char **cpp)
1.7 deraadt 1040: {
1.8 deraadt 1041: const struct op *op;
1.7 deraadt 1042: const char *cp;
1.16 sthen 1043: long val;
1044: Linetype lt, rt;
1.7 deraadt 1045:
1.16 sthen 1046: debug("eval%d", prec(ops));
1.7 deraadt 1047: cp = *cpp;
1.16 sthen 1048: lt = ops->inner(ops+1, valp, &cp);
1049: if (lt == LT_ERROR)
1050: return (LT_ERROR);
1.7 deraadt 1051: for (;;) {
1052: cp = skipcomment(cp);
1053: for (op = ops->op; op->str != NULL; op++)
1054: if (strncmp(cp, op->str, strlen(op->str)) == 0)
1055: break;
1056: if (op->str == NULL)
1057: break;
1058: cp += strlen(op->str);
1.16 sthen 1059: debug("eval%d %s", prec(ops), op->str);
1060: rt = ops->inner(ops+1, &val, &cp);
1061: if (rt == LT_ERROR)
1062: return (LT_ERROR);
1063: lt = op->fn(valp, lt, *valp, rt, val);
1.7 deraadt 1064: }
1065:
1066: *cpp = cp;
1.16 sthen 1067: debug("eval%d = %d", prec(ops), *valp);
1068: debug("eval%d lt = %s", prec(ops), linetype_name[lt]);
1069: return (lt);
1.1 deraadt 1070: }
1.7 deraadt 1071:
1.1 deraadt 1072: /*
1.7 deraadt 1073: * Evaluate the expression on a #if or #elif line. If we can work out
1074: * the result we return LT_TRUE or LT_FALSE accordingly, otherwise we
1.8 deraadt 1075: * return just a generic LT_IF.
1.1 deraadt 1076: */
1.8 deraadt 1077: static Linetype
1.7 deraadt 1078: ifeval(const char **cpp)
1079: {
1.16 sthen 1080: Linetype ret;
1081: long val = 0;
1.7 deraadt 1082:
1083: debug("eval %s", *cpp);
1.16 sthen 1084: constexpr = killconsts ? false : true;
1.8 deraadt 1085: ret = eval_table(eval_ops, &val, cpp);
1.16 sthen 1086: debug("eval = %d", val);
1087: return (constexpr ? LT_IF : ret == LT_ERROR ? LT_IF : ret);
1088: }
1089:
1090: /*
1091: * Read a line and examine its initial part to determine if it is a
1092: * preprocessor directive. Returns NULL on EOF, or a pointer to a
1093: * preprocessor directive name, or a pointer to the zero byte at the
1094: * end of the line.
1095: */
1096: static const char *
1097: skiphash(void)
1098: {
1099: const char *cp;
1100:
1101: linenum++;
1102: if (fgets(tline, MAXLINE, input) == NULL) {
1103: if (ferror(input))
1104: err(2, "can't read %s", filename);
1105: else
1106: return (NULL);
1107: }
1108: cp = skipcomment(tline);
1109: if (linestate == LS_START && *cp == '#') {
1110: linestate = LS_HASH;
1111: return (skipcomment(cp + 1));
1112: } else if (*cp == '\0') {
1113: return (cp);
1114: } else {
1115: return (skipline(cp));
1116: }
1117: }
1118:
1119: /*
1120: * Mark a line dirty and consume the rest of it, keeping track of the
1121: * lexical state.
1122: */
1123: static const char *
1124: skipline(const char *cp)
1125: {
1126: if (*cp != '\0')
1127: linestate = LS_DIRTY;
1128: while (*cp != '\0')
1129: cp = skipcomment(cp + 1);
1130: return (cp);
1.7 deraadt 1131: }
1132:
1133: /*
1.16 sthen 1134: * Skip over comments, strings, and character literals and stop at the
1135: * next character position that is not whitespace. Between calls we keep
1136: * the comment state in the global variable incomment, and we also adjust
1137: * the global variable linestate when we see a newline.
1.8 deraadt 1138: * XXX: doesn't cope with the buffer splitting inside a state transition.
1.7 deraadt 1139: */
1.8 deraadt 1140: static const char *
1.7 deraadt 1141: skipcomment(const char *cp)
1.3 deraadt 1142: {
1.8 deraadt 1143: if (text || ignoring[depth]) {
1.11 avsm 1144: for (; isspace((unsigned char)*cp); cp++)
1145: if (*cp == '\n')
1146: linestate = LS_START;
1.8 deraadt 1147: return (cp);
1148: }
1149: while (*cp != '\0')
1.16 sthen 1150: /* don't reset to LS_START after a line continuation */
1151: if (strncmp(cp, "\\\r\n", 3) == 0)
1152: cp += 3;
1153: else if (strncmp(cp, "\\\n", 2) == 0)
1.8 deraadt 1154: cp += 2;
1155: else switch (incomment) {
1156: case NO_COMMENT:
1.16 sthen 1157: if (strncmp(cp, "/\\\r\n", 4) == 0) {
1158: incomment = STARTING_COMMENT;
1159: cp += 4;
1160: } else if (strncmp(cp, "/\\\n", 3) == 0) {
1.8 deraadt 1161: incomment = STARTING_COMMENT;
1162: cp += 3;
1163: } else if (strncmp(cp, "/*", 2) == 0) {
1.3 deraadt 1164: incomment = C_COMMENT;
1.8 deraadt 1165: cp += 2;
1166: } else if (strncmp(cp, "//", 2) == 0) {
1167: incomment = CXX_COMMENT;
1168: cp += 2;
1.16 sthen 1169: } else if (strncmp(cp, "\'", 1) == 0) {
1170: incomment = CHAR_LITERAL;
1171: linestate = LS_DIRTY;
1172: cp += 1;
1173: } else if (strncmp(cp, "\"", 1) == 0) {
1174: incomment = STRING_LITERAL;
1175: linestate = LS_DIRTY;
1176: cp += 1;
1.8 deraadt 1177: } else if (strncmp(cp, "\n", 1) == 0) {
1178: linestate = LS_START;
1179: cp += 1;
1.16 sthen 1180: } else if (strchr(" \r\t", *cp) != NULL) {
1.8 deraadt 1181: cp += 1;
1182: } else
1183: return (cp);
1184: continue;
1185: case CXX_COMMENT:
1186: if (strncmp(cp, "\n", 1) == 0) {
1187: incomment = NO_COMMENT;
1188: linestate = LS_START;
1.3 deraadt 1189: }
1.8 deraadt 1190: cp += 1;
1191: continue;
1.16 sthen 1192: case CHAR_LITERAL:
1193: case STRING_LITERAL:
1194: if ((incomment == CHAR_LITERAL && cp[0] == '\'') ||
1195: (incomment == STRING_LITERAL && cp[0] == '\"')) {
1196: incomment = NO_COMMENT;
1197: cp += 1;
1198: } else if (cp[0] == '\\') {
1199: if (cp[1] == '\0')
1200: cp += 1;
1201: else
1202: cp += 2;
1203: } else if (strncmp(cp, "\n", 1) == 0) {
1204: if (incomment == CHAR_LITERAL)
1205: error("unterminated char literal");
1206: else
1207: error("unterminated string literal");
1208: } else
1209: cp += 1;
1210: continue;
1.8 deraadt 1211: case C_COMMENT:
1.16 sthen 1212: if (strncmp(cp, "*\\\r\n", 4) == 0) {
1213: incomment = FINISHING_COMMENT;
1214: cp += 4;
1215: } else if (strncmp(cp, "*\\\n", 3) == 0) {
1.8 deraadt 1216: incomment = FINISHING_COMMENT;
1217: cp += 3;
1218: } else if (strncmp(cp, "*/", 2) == 0) {
1219: incomment = NO_COMMENT;
1220: cp += 2;
1221: } else
1222: cp += 1;
1223: continue;
1224: case STARTING_COMMENT:
1225: if (*cp == '*') {
1226: incomment = C_COMMENT;
1227: cp += 1;
1228: } else if (*cp == '/') {
1.3 deraadt 1229: incomment = CXX_COMMENT;
1.8 deraadt 1230: cp += 1;
1231: } else {
1232: incomment = NO_COMMENT;
1233: linestate = LS_DIRTY;
1.3 deraadt 1234: }
1.8 deraadt 1235: continue;
1236: case FINISHING_COMMENT:
1237: if (*cp == '/') {
1238: incomment = NO_COMMENT;
1239: cp += 1;
1240: } else
1241: incomment = C_COMMENT;
1242: continue;
1243: default:
1.16 sthen 1244: abort(); /* bug */
1.3 deraadt 1245: }
1.8 deraadt 1246: return (cp);
1.1 deraadt 1247: }
1.7 deraadt 1248:
1249: /*
1.16 sthen 1250: * Skip macro arguments.
1251: */
1252: static const char *
1253: skipargs(const char *cp)
1254: {
1255: const char *ocp = cp;
1256: int level = 0;
1257: cp = skipcomment(cp);
1258: if (*cp != '(')
1259: return (cp);
1260: do {
1261: if (*cp == '(')
1262: level++;
1263: if (*cp == ')')
1264: level--;
1265: cp = skipcomment(cp+1);
1266: } while (level != 0 && *cp != '\0');
1267: if (level == 0)
1268: return (cp);
1269: else
1270: /* Rewind and re-detect the syntax error later. */
1271: return (ocp);
1272: }
1273:
1274: /*
1.7 deraadt 1275: * Skip over an identifier.
1276: */
1.8 deraadt 1277: static const char *
1.7 deraadt 1278: skipsym(const char *cp)
1279: {
1280: while (!endsym(*cp))
1281: ++cp;
1282: return (cp);
1283: }
1284:
1.1 deraadt 1285: /*
1.16 sthen 1286: * Skip whitespace and take a copy of any following identifier.
1287: */
1288: static const char *
1289: getsym(const char **cpp)
1290: {
1291: const char *cp = *cpp, *sym;
1292:
1293: cp = skipcomment(cp);
1294: cp = skipsym(sym = cp);
1295: if (cp == sym)
1296: return NULL;
1297: *cpp = cp;
1298: return (xstrdup(sym, cp));
1299: }
1300:
1301: /*
1302: * Check that s (a symbol) matches the start of t, and that the
1303: * following character in t is not a symbol character. Returns a
1304: * pointer to the following character in t if there is a match,
1305: * otherwise NULL.
1306: */
1307: static const char *
1308: matchsym(const char *s, const char *t)
1309: {
1310: while (*s != '\0' && *t != '\0')
1311: if (*s != *t)
1312: return (NULL);
1313: else
1314: ++s, ++t;
1315: if (*s == '\0' && endsym(*t))
1316: return(t);
1317: else
1318: return(NULL);
1319: }
1320:
1321: /*
1.13 jmc 1322: * Look for the symbol in the symbol table. If it is found, we return
1.8 deraadt 1323: * the symbol table index, else we return -1.
1.1 deraadt 1324: */
1.8 deraadt 1325: static int
1.16 sthen 1326: findsym(const char **strp)
1.1 deraadt 1327: {
1.16 sthen 1328: const char *str;
1.3 deraadt 1329: int symind;
1330:
1.16 sthen 1331: str = *strp;
1332: *strp = skipsym(str);
1333: if (symlist) {
1334: if (*strp == str)
1335: return (-1);
1336: if (symdepth && firstsym)
1337: printf("%s%3d", zerosyms ? "" : "\n", depth);
1338: firstsym = zerosyms = false;
1339: printf("%s%.*s%s",
1340: symdepth ? " " : "",
1341: (int)(*strp-str), str,
1342: symdepth ? "" : "\n");
1343: /* we don't care about the value of the symbol */
1344: return (0);
1345: }
1.8 deraadt 1346: for (symind = 0; symind < nsyms; ++symind) {
1.16 sthen 1347: if (matchsym(symname[symind], str) != NULL) {
1348: debugsym("findsym", symind);
1.7 deraadt 1349: return (symind);
1.3 deraadt 1350: }
1.1 deraadt 1351: }
1.8 deraadt 1352: return (-1);
1.1 deraadt 1353: }
1.7 deraadt 1354:
1.1 deraadt 1355: /*
1.16 sthen 1356: * Resolve indirect symbol values to their final definitions.
1357: */
1358: static void
1359: indirectsym(void)
1360: {
1361: const char *cp;
1362: int changed, sym, ind;
1363:
1364: do {
1365: changed = 0;
1366: for (sym = 0; sym < nsyms; ++sym) {
1367: if (value[sym] == NULL)
1368: continue;
1369: cp = value[sym];
1370: ind = findsym(&cp);
1371: if (ind == -1 || ind == sym ||
1372: *cp != '\0' ||
1373: value[ind] == NULL ||
1374: value[ind] == value[sym])
1375: continue;
1376: debugsym("indir...", sym);
1377: value[sym] = value[ind];
1378: debugsym("...ectsym", sym);
1379: changed++;
1380: }
1381: } while (changed);
1382: }
1383:
1384: /*
1385: * Add a symbol to the symbol table, specified with the format sym=val
1386: */
1387: static void
1388: addsym1(bool ignorethis, bool definethis, char *symval)
1389: {
1390: const char *sym, *val;
1391:
1392: sym = symval;
1393: val = skipsym(sym);
1394: if (definethis && *val == '=') {
1395: symval[val - sym] = '\0';
1396: val = val + 1;
1397: } else if (*val == '\0') {
1398: val = definethis ? "1" : NULL;
1399: } else {
1400: usage();
1401: }
1402: addsym2(ignorethis, sym, val);
1403: }
1404:
1405: /*
1.7 deraadt 1406: * Add a symbol to the symbol table.
1407: */
1.8 deraadt 1408: static void
1.16 sthen 1409: addsym2(bool ignorethis, const char *sym, const char *val)
1.7 deraadt 1410: {
1.16 sthen 1411: const char *cp = sym;
1.7 deraadt 1412: int symind;
1413:
1.16 sthen 1414: symind = findsym(&cp);
1.8 deraadt 1415: if (symind < 0) {
1.7 deraadt 1416: if (nsyms >= MAXSYMS)
1417: errx(2, "too many symbols");
1418: symind = nsyms++;
1419: }
1.16 sthen 1420: ignore[symind] = ignorethis;
1.7 deraadt 1421: symname[symind] = sym;
1.16 sthen 1422: value[symind] = val;
1423: debugsym("addsym", symind);
1424: }
1425:
1426: static void
1427: debugsym(const char *why, int symind)
1428: {
1429: debug("%s %s%c%s", why, symname[symind],
1430: value[symind] ? '=' : ' ',
1431: value[symind] ? value[symind] : "undef");
1432: }
1433:
1434: /*
1435: * Add symbols to the symbol table from a file containing
1436: * #define and #undef preprocessor directives.
1437: */
1438: static void
1439: defundefile(const char *fn)
1440: {
1441: filename = fn;
1442: input = fopen(fn, "rb");
1443: if (input == NULL)
1444: err(2, "can't open %s", fn);
1445: linenum = 0;
1446: while (defundef())
1447: ;
1448: if (ferror(input))
1449: err(2, "can't read %s", filename);
1450: else
1451: fclose(input);
1452: if (incomment)
1453: error("EOF in comment");
1454: }
1455:
1456: /*
1457: * Read and process one #define or #undef directive
1458: */
1459: static bool
1460: defundef(void)
1461: {
1462: const char *cp, *kw, *sym, *val, *end;
1463:
1464: cp = skiphash();
1465: if (cp == NULL)
1466: return (false);
1467: if (*cp == '\0')
1468: goto done;
1469: /* strip trailing whitespace, and do a fairly rough check to
1470: avoid unsupported multi-line preprocessor directives */
1471: end = cp + strlen(cp);
1472: while (end > tline && strchr(" \t\n\r", end[-1]) != NULL)
1473: --end;
1474: if (end > tline && end[-1] == '\\')
1475: Eioccc();
1476:
1477: kw = cp;
1478: if ((cp = matchsym("define", kw)) != NULL) {
1479: sym = getsym(&cp);
1480: if (sym == NULL)
1481: error("missing macro name in #define");
1482: if (*cp == '(') {
1483: val = "1";
1484: } else {
1485: cp = skipcomment(cp);
1486: val = (cp < end) ? xstrdup(cp, end) : "";
1487: }
1488: debug("#define");
1489: addsym2(false, sym, val);
1490: } else if ((cp = matchsym("undef", kw)) != NULL) {
1491: sym = getsym(&cp);
1492: if (sym == NULL)
1493: error("missing macro name in #undef");
1494: cp = skipcomment(cp);
1495: debug("#undef");
1496: addsym2(false, sym, NULL);
1.7 deraadt 1497: } else {
1.16 sthen 1498: error("unrecognized preprocessor directive");
1.7 deraadt 1499: }
1.16 sthen 1500: skipline(cp);
1501: done:
1502: debug("parser line %d state %s comment %s line", linenum,
1503: comment_name[incomment], linestate_name[linestate]);
1504: return (true);
1505: }
1506:
1507: /*
1508: * Concatenate two strings into new memory, checking for failure.
1509: */
1510: static char *
1511: astrcat(const char *s1, const char *s2)
1512: {
1513: char *s;
1514: int len;
1515: size_t size;
1516:
1517: len = snprintf(NULL, 0, "%s%s", s1, s2);
1518: if (len < 0)
1519: err(2, "snprintf");
1520: size = (size_t)len + 1;
1.20 ! deraadt 1521: s = malloc(size);
1.16 sthen 1522: if (s == NULL)
1523: err(2, "malloc");
1524: snprintf(s, size, "%s%s", s1, s2);
1525: return (s);
1.7 deraadt 1526: }
1527:
1528: /*
1.16 sthen 1529: * Duplicate a segment of a string, checking for failure.
1.1 deraadt 1530: */
1.16 sthen 1531: static const char *
1532: xstrdup(const char *start, const char *end)
1.3 deraadt 1533: {
1.16 sthen 1534: size_t n;
1535: char *s;
1536:
1537: if (end < start) abort(); /* bug */
1538: n = (size_t)(end - start) + 1;
1539: s = malloc(n);
1540: if (s == NULL)
1541: err(2, "malloc");
1542: snprintf(s, n, "%s", start);
1543: return (s);
1.1 deraadt 1544: }
1545:
1.7 deraadt 1546: /*
1.8 deraadt 1547: * Diagnostics.
1.7 deraadt 1548: */
1.8 deraadt 1549: static void
1.7 deraadt 1550: debug(const char *msg, ...)
1.1 deraadt 1551: {
1.7 deraadt 1552: va_list ap;
1553:
1554: if (debugging) {
1555: va_start(ap, msg);
1556: vwarnx(msg, ap);
1557: va_end(ap);
1558: }
1.1 deraadt 1559: }
1560:
1.8 deraadt 1561: static void
1562: error(const char *msg)
1.7 deraadt 1563: {
1.8 deraadt 1564: if (depth == 0)
1.9 deraadt 1565: warnx("%s: %d: %s", filename, linenum, msg);
1.7 deraadt 1566: else
1.9 deraadt 1567: warnx("%s: %d: %s (#if line %d depth %d)",
1.8 deraadt 1568: filename, linenum, msg, stifline[depth], depth);
1.16 sthen 1569: closeio();
1.9 deraadt 1570: errx(2, "output may be truncated");
1.1 deraadt 1571: }