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