2 * Copyright (c) 2002 - 2011 Tony Finch <dot@dotat.at>
4 * Redistribution and use in source and binary forms, with or without
5 * modification, are permitted provided that the following conditions
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.
13 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
14 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
15 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
16 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
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
27 * unifdef - remove ifdef'ed lines
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.
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
42 * These require better buffer handling, which would also make
43 * it possible to handle all "dodgy" directives correctly.
46 #include <sys/types.h>
59 const char copyright
[] =
60 "@(#) $Version: unifdef-2.5 $\n"
61 "@(#) $Author: Tony Finch (dot@dotat.at) $\n"
62 "@(#) $URL: http://dotat.at/prog/unifdef $\n"
65 /* types of input lines: */
67 LT_TRUEI
, /* a true #if with ignore flag */
68 LT_FALSEI
, /* a false #if with ignore flag */
69 LT_IF
, /* an unknown #if */
70 LT_TRUE
, /* a true #if */
71 LT_FALSE
, /* a false #if */
72 LT_ELIF
, /* an unknown #elif */
73 LT_ELTRUE
, /* a true #elif */
74 LT_ELFALSE
, /* a false #elif */
76 LT_ENDIF
, /* #endif */
77 LT_DODGY
, /* flag: directive is not on one line */
78 LT_DODGY_LAST
= LT_DODGY
+ LT_ENDIF
,
79 LT_PLAIN
, /* ordinary line */
80 LT_EOF
, /* end of file */
81 LT_ERROR
, /* unevaluable #if */
85 static char const * const linetype_name
[] = {
86 "TRUEI", "FALSEI", "IF", "TRUE", "FALSE",
87 "ELIF", "ELTRUE", "ELFALSE", "ELSE", "ENDIF",
88 "DODGY TRUEI", "DODGY FALSEI",
89 "DODGY IF", "DODGY TRUE", "DODGY FALSE",
90 "DODGY ELIF", "DODGY ELTRUE", "DODGY ELFALSE",
91 "DODGY ELSE", "DODGY ENDIF",
92 "PLAIN", "EOF", "ERROR"
95 /* state of #if processing */
98 IS_FALSE_PREFIX
, /* false #if followed by false #elifs */
99 IS_TRUE_PREFIX
, /* first non-false #(el)if is true */
100 IS_PASS_MIDDLE
, /* first non-false #(el)if is unknown */
101 IS_FALSE_MIDDLE
, /* a false #elif after a pass state */
102 IS_TRUE_MIDDLE
, /* a true #elif after a pass state */
103 IS_PASS_ELSE
, /* an else after a pass state */
104 IS_FALSE_ELSE
, /* an else after a true state */
105 IS_TRUE_ELSE
, /* an else after only false states */
106 IS_FALSE_TRAILER
, /* #elifs after a true are false */
110 static char const * const ifstate_name
[] = {
111 "OUTSIDE", "FALSE_PREFIX", "TRUE_PREFIX",
112 "PASS_MIDDLE", "FALSE_MIDDLE", "TRUE_MIDDLE",
113 "PASS_ELSE", "FALSE_ELSE", "TRUE_ELSE",
117 /* state of comment parser */
119 NO_COMMENT
= false, /* outside a comment */
120 C_COMMENT
, /* in a comment like this one */
121 CXX_COMMENT
, /* between // and end of line */
122 STARTING_COMMENT
, /* just after slash-backslash-newline */
123 FINISHING_COMMENT
, /* star-backslash-newline in a C comment */
124 CHAR_LITERAL
, /* inside '' */
125 STRING_LITERAL
/* inside "" */
128 static char const * const comment_name
[] = {
129 "NO", "C", "CXX", "STARTING", "FINISHING", "CHAR", "STRING"
132 /* state of preprocessor line parser */
134 LS_START
, /* only space and comments on this line */
135 LS_HASH
, /* only space, comments, and a hash */
136 LS_DIRTY
/* this line can't be a preprocessor line */
139 static char const * const linestate_name
[] = {
140 "START", "HASH", "DIRTY"
144 * Minimum translation limits from ISO/IEC 9899:1999 5.2.4.1
146 #define MAXDEPTH 64 /* maximum #if nesting */
147 #define MAXLINE 4096 /* maximum length of line */
148 #define MAXSYMS 4096 /* maximum number of symbols */
151 * Sometimes when editing a keyword the replacement text is longer, so
152 * we leave some space at the end of the tline buffer to accommodate this.
157 * For temporary filenames
159 #define TEMPLATE "unifdef.XXXXXX"
165 static bool compblank
; /* -B: compress blank lines */
166 static bool lnblank
; /* -b: blank deleted lines */
167 static bool complement
; /* -c: do the complement */
168 static bool debugging
; /* -d: debugging reports */
169 static bool iocccok
; /* -e: fewer IOCCC errors */
170 static bool strictlogic
; /* -K: keep ambiguous #ifs */
171 static bool killconsts
; /* -k: eval constant #ifs */
172 static bool lnnum
; /* -n: add #line directives */
173 static bool symlist
; /* -s: output symbol list */
174 static bool symdepth
; /* -S: output symbol depth */
175 static bool text
; /* -t: this is a text file */
177 static const char *symname
[MAXSYMS
]; /* symbol name */
178 static const char *value
[MAXSYMS
]; /* -Dsym=value */
179 static bool ignore
[MAXSYMS
]; /* -iDsym or -iUsym */
180 static int nsyms
; /* number of symbols */
182 static FILE *input
; /* input file pointer */
183 static const char *filename
; /* input file name */
184 static int linenum
; /* current line number */
185 static FILE *output
; /* output file pointer */
186 static const char *ofilename
; /* output file name */
187 static bool overwriting
; /* output overwrites input */
188 static char tempname
[FILENAME_MAX
]; /* used when overwriting */
190 static char tline
[MAXLINE
+EDITSLOP
];/* input buffer plus space */
191 static char *keyword
; /* used for editing #elif's */
193 static const char *newline
; /* input file format */
194 static const char newline_unix
[] = "\n";
195 static const char newline_crlf
[] = "\r\n";
197 static Comment_state incomment
; /* comment parser state */
198 static Line_state linestate
; /* #if line parser state */
199 static Ifstate ifstate
[MAXDEPTH
]; /* #if processor state */
200 static bool ignoring
[MAXDEPTH
]; /* ignore comments state */
201 static int stifline
[MAXDEPTH
]; /* start of current #if */
202 static int depth
; /* current #if nesting */
203 static int delcount
; /* count of deleted lines */
204 static unsigned blankcount
; /* count of blank lines */
205 static unsigned blankmax
; /* maximum recent blankcount */
206 static bool constexpr; /* constant #if expression */
207 static bool zerosyms
= true; /* to format symdepth output */
208 static bool firstsym
; /* ditto */
210 static int exitstat
; /* program exit status */
212 static void addsym(bool, bool, char *);
213 static void closeout(void);
214 static void debug(const char *, ...);
215 static void done(void);
216 static void error(const char *);
217 static int findsym(const char *);
218 static void flushline(bool);
219 static Linetype
parseline(void);
220 static Linetype
ifeval(const char **);
221 static void ignoreoff(void);
222 static void ignoreon(void);
223 static void keywordedit(const char *);
224 static void nest(void);
225 static void process(void);
226 static const char *skipargs(const char *);
227 static const char *skipcomment(const char *);
228 static const char *skipsym(const char *);
229 static void state(Ifstate
);
230 static int strlcmp(const char *, const char *, size_t);
231 static void unnest(void);
232 static void usage(void);
233 static void version(void);
235 #define endsym(c) (!isalnum((unsigned char)c) && c != '_')
241 main(int argc
, char *argv
[])
245 while ((opt
= getopt(argc
, argv
, "i:D:U:I:o:bBcdeKklnsStV")) != -1)
247 case 'i': /* treat stuff controlled by these symbols as text */
249 * For strict backwards-compatibility the U or D
250 * should be immediately after the -i but it doesn't
251 * matter much if we relax that requirement.
255 addsym(true, true, optarg
);
257 addsym(true, false, optarg
);
261 case 'D': /* define a symbol */
262 addsym(false, true, optarg
);
264 case 'U': /* undef a symbol */
265 addsym(false, false, optarg
);
267 case 'I': /* no-op for compatibility with cpp */
269 case 'b': /* blank deleted lines instead of omitting them */
270 case 'l': /* backwards compatibility */
273 case 'B': /* compress blank lines around removed section */
276 case 'c': /* treat -D as -U and vice versa */
282 case 'e': /* fewer errors from dodgy lines */
285 case 'K': /* keep ambiguous #ifs */
288 case 'k': /* process constant #ifs */
291 case 'n': /* add #line directive after deleted lines */
294 case 'o': /* output to a file */
297 case 's': /* only output list of symbols that control #ifs */
300 case 'S': /* list symbols with their nesting depth */
301 symlist
= symdepth
= true;
303 case 't': /* don't parse C comments */
306 case 'V': /* print version */
313 if (compblank
&& lnblank
)
314 errx(2, "-B and -b are mutually exclusive");
316 errx(2, "can only do one file");
317 } else if (argc
== 1 && strcmp(*argv
, "-") != 0) {
319 input
= fopen(filename
, "rb");
321 err(2, "can't open %s", filename
);
323 filename
= "[stdin]";
326 if (ofilename
== NULL
) {
327 ofilename
= "[stdout]";
330 struct stat ist
, ost
;
331 if (stat(ofilename
, &ost
) == 0 &&
332 fstat(fileno(input
), &ist
) == 0)
333 overwriting
= (ist
.st_dev
== ost
.st_dev
334 && ist
.st_ino
== ost
.st_ino
);
339 dirsep
= strrchr(ofilename
, '/');
341 snprintf(tempname
, sizeof(tempname
),
343 (int)(dirsep
- ofilename
), ofilename
);
345 snprintf(tempname
, sizeof(tempname
),
347 ofd
= mkstemp(tempname
);
349 output
= fdopen(ofd
, "wb+");
351 err(2, "can't create temporary file");
352 fchmod(ofd
, ist
.st_mode
& (S_IRWXU
|S_IRWXG
|S_IRWXO
));
354 output
= fopen(ofilename
, "wb");
356 err(2, "can't open %s", ofilename
);
366 const char *c
= copyright
;
380 fprintf(stderr
, "usage: unifdef [-bBcdeKknsStV] [-Ipath]"
381 " [-Dsym[=val]] [-Usym] [-iDsym[=val]] [-iUsym] ... [file]\n");
386 * A state transition function alters the global #if processing state
387 * in a particular way. The table below is indexed by the current
388 * processing state and the type of the current line.
390 * Nesting is handled by keeping a stack of states; some transition
391 * functions increase or decrease the depth. They also maintain the
392 * ignore state on a stack. In some complicated cases they have to
393 * alter the preprocessor directive, as follows.
395 * When we have processed a group that starts off with a known-false
396 * #if/#elif sequence (which has therefore been deleted) followed by a
397 * #elif that we don't understand and therefore must keep, we edit the
398 * latter into a #if to keep the nesting correct. We use memcpy() to
399 * overwrite the 4 byte token "elif" with "if " without a '\0' byte.
401 * When we find a true #elif in a group, the following block will
402 * always be kept and the rest of the sequence after the next #elif or
403 * #else will be discarded. We edit the #elif into a #else and the
404 * following directive to #endif since this has the desired behaviour.
406 * "Dodgy" directives are split across multiple lines, the most common
407 * example being a multi-line comment hanging off the right of the
408 * directive. We can handle them correctly only if there is no change
409 * from printing to dropping (or vice versa) caused by that directive.
410 * If the directive is the first of a group we have a choice between
411 * failing with an error, or passing it through unchanged instead of
412 * evaluating it. The latter is not the default to avoid questions from
413 * users about unifdef unexpectedly leaving behind preprocessor directives.
415 typedef void state_fn(void);
417 /* report an error */
418 static void Eelif (void) { error("Inappropriate #elif"); }
419 static void Eelse (void) { error("Inappropriate #else"); }
420 static void Eendif(void) { error("Inappropriate #endif"); }
421 static void Eeof (void) { error("Premature EOF"); }
422 static void Eioccc(void) { error("Obfuscated preprocessor control line"); }
423 /* plain line handling */
424 static void print (void) { flushline(true); }
425 static void drop (void) { flushline(false); }
426 /* output lacks group's start line */
427 static void Strue (void) { drop(); ignoreoff(); state(IS_TRUE_PREFIX
); }
428 static void Sfalse(void) { drop(); ignoreoff(); state(IS_FALSE_PREFIX
); }
429 static void Selse (void) { drop(); state(IS_TRUE_ELSE
); }
430 /* print/pass this block */
431 static void Pelif (void) { print(); ignoreoff(); state(IS_PASS_MIDDLE
); }
432 static void Pelse (void) { print(); state(IS_PASS_ELSE
); }
433 static void Pendif(void) { print(); unnest(); }
434 /* discard this block */
435 static void Dfalse(void) { drop(); ignoreoff(); state(IS_FALSE_TRAILER
); }
436 static void Delif (void) { drop(); ignoreoff(); state(IS_FALSE_MIDDLE
); }
437 static void Delse (void) { drop(); state(IS_FALSE_ELSE
); }
438 static void Dendif(void) { drop(); unnest(); }
439 /* first line of group */
440 static void Fdrop (void) { nest(); Dfalse(); }
441 static void Fpass (void) { nest(); Pelif(); }
442 static void Ftrue (void) { nest(); Strue(); }
443 static void Ffalse(void) { nest(); Sfalse(); }
444 /* variable pedantry for obfuscated lines */
445 static void Oiffy (void) { if (!iocccok
) Eioccc(); Fpass(); ignoreon(); }
446 static void Oif (void) { if (!iocccok
) Eioccc(); Fpass(); }
447 static void Oelif (void) { if (!iocccok
) Eioccc(); Pelif(); }
448 /* ignore comments in this block */
449 static void Idrop (void) { Fdrop(); ignoreon(); }
450 static void Itrue (void) { Ftrue(); ignoreon(); }
451 static void Ifalse(void) { Ffalse(); ignoreon(); }
452 /* modify this line */
453 static void Mpass (void) { memcpy(keyword
, "if ", 4); Pelif(); }
454 static void Mtrue (void) { keywordedit("else"); state(IS_TRUE_MIDDLE
); }
455 static void Melif (void) { keywordedit("endif"); state(IS_FALSE_TRAILER
); }
456 static void Melse (void) { keywordedit("endif"); state(IS_FALSE_ELSE
); }
458 static state_fn
* const trans_table
[IS_COUNT
][LT_COUNT
] = {
460 { Itrue
, Ifalse
,Fpass
, Ftrue
, Ffalse
,Eelif
, Eelif
, Eelif
, Eelse
, Eendif
,
461 Oiffy
, Oiffy
, Fpass
, Oif
, Oif
, Eelif
, Eelif
, Eelif
, Eelse
, Eendif
,
462 print
, done
, abort
},
463 /* IS_FALSE_PREFIX */
464 { Idrop
, Idrop
, Fdrop
, Fdrop
, Fdrop
, Mpass
, Strue
, Sfalse
,Selse
, Dendif
,
465 Idrop
, Idrop
, Fdrop
, Fdrop
, Fdrop
, Mpass
, Eioccc
,Eioccc
,Eioccc
,Eioccc
,
468 { Itrue
, Ifalse
,Fpass
, Ftrue
, Ffalse
,Dfalse
,Dfalse
,Dfalse
,Delse
, Dendif
,
469 Oiffy
, Oiffy
, Fpass
, Oif
, Oif
, Eioccc
,Eioccc
,Eioccc
,Eioccc
,Eioccc
,
470 print
, Eeof
, abort
},
472 { Itrue
, Ifalse
,Fpass
, Ftrue
, Ffalse
,Pelif
, Mtrue
, Delif
, Pelse
, Pendif
,
473 Oiffy
, Oiffy
, Fpass
, Oif
, Oif
, Pelif
, Oelif
, Oelif
, Pelse
, Pendif
,
474 print
, Eeof
, abort
},
475 /* IS_FALSE_MIDDLE */
476 { Idrop
, Idrop
, Fdrop
, Fdrop
, Fdrop
, Pelif
, Mtrue
, Delif
, Pelse
, Pendif
,
477 Idrop
, Idrop
, Fdrop
, Fdrop
, Fdrop
, Eioccc
,Eioccc
,Eioccc
,Eioccc
,Eioccc
,
480 { Itrue
, Ifalse
,Fpass
, Ftrue
, Ffalse
,Melif
, Melif
, Melif
, Melse
, Pendif
,
481 Oiffy
, Oiffy
, Fpass
, Oif
, Oif
, Eioccc
,Eioccc
,Eioccc
,Eioccc
,Pendif
,
482 print
, Eeof
, abort
},
484 { Itrue
, Ifalse
,Fpass
, Ftrue
, Ffalse
,Eelif
, Eelif
, Eelif
, Eelse
, Pendif
,
485 Oiffy
, Oiffy
, Fpass
, Oif
, Oif
, Eelif
, Eelif
, Eelif
, Eelse
, Pendif
,
486 print
, Eeof
, abort
},
488 { Idrop
, Idrop
, Fdrop
, Fdrop
, Fdrop
, Eelif
, Eelif
, Eelif
, Eelse
, Dendif
,
489 Idrop
, Idrop
, Fdrop
, Fdrop
, Fdrop
, Eelif
, Eelif
, Eelif
, Eelse
, Eioccc
,
492 { Itrue
, Ifalse
,Fpass
, Ftrue
, Ffalse
,Eelif
, Eelif
, Eelif
, Eelse
, Dendif
,
493 Oiffy
, Oiffy
, Fpass
, Oif
, Oif
, Eelif
, Eelif
, Eelif
, Eelse
, Eioccc
,
494 print
, Eeof
, abort
},
495 /* IS_FALSE_TRAILER */
496 { Idrop
, Idrop
, Fdrop
, Fdrop
, Fdrop
, Dfalse
,Dfalse
,Dfalse
,Delse
, Dendif
,
497 Idrop
, Idrop
, Fdrop
, Fdrop
, Fdrop
, Dfalse
,Dfalse
,Dfalse
,Delse
, Eioccc
,
499 /*TRUEI FALSEI IF TRUE FALSE ELIF ELTRUE ELFALSE ELSE ENDIF
500 TRUEI FALSEI IF TRUE FALSE ELIF ELTRUE ELFALSE ELSE ENDIF (DODGY)
505 * State machine utility functions
512 ignoring
[depth
] = ignoring
[depth
-1];
517 ignoring
[depth
] = true;
520 keywordedit(const char *replacement
)
522 snprintf(keyword
, tline
+ sizeof(tline
) - keyword
,
523 "%s%s", replacement
, newline
);
529 if (depth
> MAXDEPTH
-1)
531 if (depth
== MAXDEPTH
-1)
532 error("Too many levels of nesting");
534 stifline
[depth
] = linenum
;
550 * Write a line to the output or not, according to command line options.
557 if (keep
^ complement
) {
558 bool blankline
= tline
[strspn(tline
, " \t\r\n")] == '\0';
559 if (blankline
&& compblank
&& blankcount
!= blankmax
) {
563 if (lnnum
&& delcount
> 0)
564 printf("#line %d%s", linenum
, newline
);
565 fputs(tline
, output
);
567 blankmax
= blankcount
= blankline
? blankcount
+ 1 : 0;
571 fputs(newline
, output
);
581 * The driver for the state machine.
586 /* When compressing blank lines, act as if the file
587 is preceded by a large number of blank lines. */
588 blankmax
= blankcount
= 1000;
590 Linetype lineval
= parseline();
591 trans_table
[ifstate
[depth
]][lineval
]();
592 debug("process line %d %s -> %s depth %d",
593 linenum
, linetype_name
[lineval
],
594 ifstate_name
[ifstate
[depth
]], depth
);
599 * Flush the output and handle errors.
604 if (symdepth
&& !zerosyms
)
606 if (fclose(output
) == EOF
) {
607 warn("couldn't write to %s", ofilename
);
610 errx(2, "%s unchanged", filename
);
624 error("EOF in comment");
626 if (overwriting
&& rename(tempname
, ofilename
) == -1) {
627 warn("couldn't rename temporary file");
629 errx(2, "%s unchanged", ofilename
);
635 * Parse a line and determine its type. We keep the preprocessor line
636 * parser state between calls in the global variable linestate, with
637 * help from skipcomment().
646 Comment_state wascomment
;
649 if (fgets(tline
, MAXLINE
, input
) == NULL
)
651 if (newline
== NULL
) {
652 if (strrchr(tline
, '\n') == strrchr(tline
, '\r') + 1)
653 newline
= newline_crlf
;
655 newline
= newline_unix
;
658 wascomment
= incomment
;
659 cp
= skipcomment(tline
);
660 if (linestate
== LS_START
) {
664 cp
= skipcomment(cp
+ 1);
665 } else if (*cp
!= '\0')
666 linestate
= LS_DIRTY
;
668 if (!incomment
&& linestate
== LS_HASH
) {
669 keyword
= tline
+ (cp
- tline
);
671 kwlen
= cp
- keyword
;
672 /* no way can we deal with a continuation inside a keyword */
673 if (strncmp(cp
, "\\\r\n", 3) == 0 ||
674 strncmp(cp
, "\\\n", 2) == 0)
676 if (strlcmp("ifdef", keyword
, kwlen
) == 0 ||
677 strlcmp("ifndef", keyword
, kwlen
) == 0) {
678 cp
= skipcomment(cp
);
679 if ((cursym
= findsym(cp
)) < 0)
682 retval
= (keyword
[2] == 'n')
683 ? LT_FALSE
: LT_TRUE
;
684 if (value
[cursym
] == NULL
)
685 retval
= (retval
== LT_TRUE
)
686 ? LT_FALSE
: LT_TRUE
;
688 retval
= (retval
== LT_TRUE
)
689 ? LT_TRUEI
: LT_FALSEI
;
692 } else if (strlcmp("if", keyword
, kwlen
) == 0)
693 retval
= ifeval(&cp
);
694 else if (strlcmp("elif", keyword
, kwlen
) == 0)
695 retval
= ifeval(&cp
) - LT_IF
+ LT_ELIF
;
696 else if (strlcmp("else", keyword
, kwlen
) == 0)
698 else if (strlcmp("endif", keyword
, kwlen
) == 0)
701 linestate
= LS_DIRTY
;
704 cp
= skipcomment(cp
);
706 linestate
= LS_DIRTY
;
707 if (retval
== LT_TRUE
|| retval
== LT_FALSE
||
708 retval
== LT_TRUEI
|| retval
== LT_FALSEI
)
710 if (retval
== LT_ELTRUE
|| retval
== LT_ELFALSE
)
713 if (retval
!= LT_PLAIN
&& (wascomment
|| incomment
)) {
716 linestate
= LS_DIRTY
;
718 /* skipcomment normally changes the state, except
719 if the last line of the file lacks a newline, or
720 if there is too much whitespace in a directive */
721 if (linestate
== LS_HASH
) {
722 size_t len
= cp
- tline
;
723 if (fgets(tline
+ len
, MAXLINE
- len
, input
) == NULL
) {
724 /* append the missing newline */
725 strcpy(tline
+ len
, newline
);
726 cp
+= strlen(newline
);
727 linestate
= LS_START
;
729 linestate
= LS_DIRTY
;
733 if (linestate
== LS_DIRTY
) {
735 cp
= skipcomment(cp
+ 1);
737 debug("parser line %d state %s comment %s line", linenum
,
738 comment_name
[incomment
], linestate_name
[linestate
]);
743 * These are the binary operators that are supported by the expression
746 static Linetype
op_strict(int *p
, int v
, Linetype at
, Linetype bt
) {
747 if(at
== LT_IF
|| bt
== LT_IF
) return (LT_IF
);
748 return (*p
= v
, v
? LT_TRUE
: LT_FALSE
);
750 static Linetype
op_lt(int *p
, Linetype at
, int a
, Linetype bt
, int b
) {
751 return op_strict(p
, a
< b
, at
, bt
);
753 static Linetype
op_gt(int *p
, Linetype at
, int a
, Linetype bt
, int b
) {
754 return op_strict(p
, a
> b
, at
, bt
);
756 static Linetype
op_le(int *p
, Linetype at
, int a
, Linetype bt
, int b
) {
757 return op_strict(p
, a
<= b
, at
, bt
);
759 static Linetype
op_ge(int *p
, Linetype at
, int a
, Linetype bt
, int b
) {
760 return op_strict(p
, a
>= b
, at
, bt
);
762 static Linetype
op_eq(int *p
, Linetype at
, int a
, Linetype bt
, int b
) {
763 return op_strict(p
, a
== b
, at
, bt
);
765 static Linetype
op_ne(int *p
, Linetype at
, int a
, Linetype bt
, int b
) {
766 return op_strict(p
, a
!= b
, at
, bt
);
768 static Linetype
op_or(int *p
, Linetype at
, int a
, Linetype bt
, int b
) {
769 if (!strictlogic
&& (at
== LT_TRUE
|| bt
== LT_TRUE
))
770 return (*p
= 1, LT_TRUE
);
771 return op_strict(p
, a
|| b
, at
, bt
);
773 static Linetype
op_and(int *p
, Linetype at
, int a
, Linetype bt
, int b
) {
774 if (!strictlogic
&& (at
== LT_FALSE
|| bt
== LT_FALSE
))
775 return (*p
= 0, LT_FALSE
);
776 return op_strict(p
, a
&& b
, at
, bt
);
780 * An evaluation function takes three arguments, as follows: (1) a pointer to
781 * an element of the precedence table which lists the operators at the current
782 * level of precedence; (2) a pointer to an integer which will receive the
783 * value of the expression; and (3) a pointer to a char* that points to the
784 * expression to be evaluated and that is updated to the end of the expression
785 * when evaluation is complete. The function returns LT_FALSE if the value of
786 * the expression is zero, LT_TRUE if it is non-zero, LT_IF if the expression
787 * depends on an unknown symbol, or LT_ERROR if there is a parse failure.
791 typedef Linetype
eval_fn(const struct ops
*, int *, const char **);
793 static eval_fn eval_table
, eval_unary
;
796 * The precedence table. Expressions involving binary operators are evaluated
797 * in a table-driven way by eval_table. When it evaluates a subexpression it
798 * calls the inner function with its first argument pointing to the next
799 * element of the table. Innermost expressions have special non-table-driven
802 static const struct ops
{
806 Linetype (*fn
)(int *, Linetype
, int, Linetype
, int);
809 { eval_table
, { { "||", op_or
} } },
810 { eval_table
, { { "&&", op_and
} } },
811 { eval_table
, { { "==", op_eq
},
813 { eval_unary
, { { "<=", op_le
},
820 * Function for evaluating the innermost parts of expressions,
821 * viz. !expr (expr) number defined(symbol) symbol
822 * We reset the constexpr flag in the last two cases.
825 eval_unary(const struct ops
*ops
, int *valp
, const char **cpp
)
833 cp
= skipcomment(*cpp
);
835 debug("eval%d !", ops
- eval_ops
);
837 lt
= eval_unary(ops
, valp
, &cp
);
842 lt
= *valp
? LT_TRUE
: LT_FALSE
;
844 } else if (*cp
== '(') {
846 debug("eval%d (", ops
- eval_ops
);
847 lt
= eval_table(eval_ops
, valp
, &cp
);
850 cp
= skipcomment(cp
);
853 } else if (isdigit((unsigned char)*cp
)) {
854 debug("eval%d number", ops
- eval_ops
);
855 *valp
= strtol(cp
, &ep
, 0);
858 lt
= *valp
? LT_TRUE
: LT_FALSE
;
860 } else if (strncmp(cp
, "defined", 7) == 0 && endsym(cp
[7])) {
861 cp
= skipcomment(cp
+7);
862 debug("eval%d defined", ops
- eval_ops
);
864 cp
= skipcomment(cp
+1);
873 *valp
= (value
[sym
] != NULL
);
874 lt
= *valp
? LT_TRUE
: LT_FALSE
;
877 cp
= skipcomment(cp
);
878 if (defparen
&& *cp
++ != ')')
881 } else if (!endsym(*cp
)) {
882 debug("eval%d symbol", ops
- eval_ops
);
888 } else if (value
[sym
] == NULL
) {
892 *valp
= strtol(value
[sym
], &ep
, 0);
893 if (*ep
!= '\0' || ep
== value
[sym
])
895 lt
= *valp
? LT_TRUE
: LT_FALSE
;
900 debug("eval%d bad expr", ops
- eval_ops
);
905 debug("eval%d = %d", ops
- eval_ops
, *valp
);
910 * Table-driven evaluation of binary operators.
913 eval_table(const struct ops
*ops
, int *valp
, const char **cpp
)
920 debug("eval%d", ops
- eval_ops
);
922 lt
= ops
->inner(ops
+1, valp
, &cp
);
926 cp
= skipcomment(cp
);
927 for (op
= ops
->op
; op
->str
!= NULL
; op
++)
928 if (strncmp(cp
, op
->str
, strlen(op
->str
)) == 0)
932 cp
+= strlen(op
->str
);
933 debug("eval%d %s", ops
- eval_ops
, op
->str
);
934 rt
= ops
->inner(ops
+1, &val
, &cp
);
937 lt
= op
->fn(valp
, lt
, *valp
, rt
, val
);
941 debug("eval%d = %d", ops
- eval_ops
, *valp
);
942 debug("eval%d lt = %s", ops
- eval_ops
, linetype_name
[lt
]);
947 * Evaluate the expression on a #if or #elif line. If we can work out
948 * the result we return LT_TRUE or LT_FALSE accordingly, otherwise we
949 * return just a generic LT_IF.
952 ifeval(const char **cpp
)
957 debug("eval %s", *cpp
);
958 constexpr = killconsts
? false : true;
959 ret
= eval_table(eval_ops
, &val
, cpp
);
960 debug("eval = %d", val
);
961 return (constexpr ? LT_IF
: ret
== LT_ERROR
? LT_IF
: ret
);
965 * Skip over comments, strings, and character literals and stop at the
966 * next character position that is not whitespace. Between calls we keep
967 * the comment state in the global variable incomment, and we also adjust
968 * the global variable linestate when we see a newline.
969 * XXX: doesn't cope with the buffer splitting inside a state transition.
972 skipcomment(const char *cp
)
974 if (text
|| ignoring
[depth
]) {
975 for (; isspace((unsigned char)*cp
); cp
++)
977 linestate
= LS_START
;
981 /* don't reset to LS_START after a line continuation */
982 if (strncmp(cp
, "\\\r\n", 3) == 0)
984 else if (strncmp(cp
, "\\\n", 2) == 0)
986 else switch (incomment
) {
988 if (strncmp(cp
, "/\\\r\n", 4) == 0) {
989 incomment
= STARTING_COMMENT
;
991 } else if (strncmp(cp
, "/\\\n", 3) == 0) {
992 incomment
= STARTING_COMMENT
;
994 } else if (strncmp(cp
, "/*", 2) == 0) {
995 incomment
= C_COMMENT
;
997 } else if (strncmp(cp
, "//", 2) == 0) {
998 incomment
= CXX_COMMENT
;
1000 } else if (strncmp(cp
, "\'", 1) == 0) {
1001 incomment
= CHAR_LITERAL
;
1002 linestate
= LS_DIRTY
;
1004 } else if (strncmp(cp
, "\"", 1) == 0) {
1005 incomment
= STRING_LITERAL
;
1006 linestate
= LS_DIRTY
;
1008 } else if (strncmp(cp
, "\n", 1) == 0) {
1009 linestate
= LS_START
;
1011 } else if (strchr(" \r\t", *cp
) != NULL
) {
1017 if (strncmp(cp
, "\n", 1) == 0) {
1018 incomment
= NO_COMMENT
;
1019 linestate
= LS_START
;
1024 case STRING_LITERAL
:
1025 if ((incomment
== CHAR_LITERAL
&& cp
[0] == '\'') ||
1026 (incomment
== STRING_LITERAL
&& cp
[0] == '\"')) {
1027 incomment
= NO_COMMENT
;
1029 } else if (cp
[0] == '\\') {
1034 } else if (strncmp(cp
, "\n", 1) == 0) {
1035 if (incomment
== CHAR_LITERAL
)
1036 error("unterminated char literal");
1038 error("unterminated string literal");
1043 if (strncmp(cp
, "*\\\r\n", 4) == 0) {
1044 incomment
= FINISHING_COMMENT
;
1046 } else if (strncmp(cp
, "*\\\n", 3) == 0) {
1047 incomment
= FINISHING_COMMENT
;
1049 } else if (strncmp(cp
, "*/", 2) == 0) {
1050 incomment
= NO_COMMENT
;
1055 case STARTING_COMMENT
:
1057 incomment
= C_COMMENT
;
1059 } else if (*cp
== '/') {
1060 incomment
= CXX_COMMENT
;
1063 incomment
= NO_COMMENT
;
1064 linestate
= LS_DIRTY
;
1067 case FINISHING_COMMENT
:
1069 incomment
= NO_COMMENT
;
1072 incomment
= C_COMMENT
;
1081 * Skip macro arguments.
1084 skipargs(const char *cp
)
1086 const char *ocp
= cp
;
1088 cp
= skipcomment(cp
);
1096 cp
= skipcomment(cp
+1);
1097 } while (level
!= 0 && *cp
!= '\0');
1101 /* Rewind and re-detect the syntax error later. */
1106 * Skip over an identifier.
1109 skipsym(const char *cp
)
1111 while (!endsym(*cp
))
1117 * Look for the symbol in the symbol table. If it is found, we return
1118 * the symbol table index, else we return -1.
1121 findsym(const char *str
)
1130 if (symdepth
&& firstsym
)
1131 printf("%s%3d", zerosyms
? "" : "\n", depth
);
1132 firstsym
= zerosyms
= false;
1134 symdepth
? " " : "",
1136 symdepth
? "" : "\n");
1137 /* we don't care about the value of the symbol */
1140 for (symind
= 0; symind
< nsyms
; ++symind
) {
1141 if (strlcmp(symname
[symind
], str
, cp
-str
) == 0) {
1142 debug("findsym %s %s", symname
[symind
],
1143 value
[symind
] ? value
[symind
] : "");
1151 * Add a symbol to the symbol table.
1154 addsym(bool ignorethis
, bool definethis
, char *sym
)
1159 symind
= findsym(sym
);
1161 if (nsyms
>= MAXSYMS
)
1162 errx(2, "too many symbols");
1165 symname
[symind
] = sym
;
1166 ignore
[symind
] = ignorethis
;
1167 val
= sym
+ (skipsym(sym
) - sym
);
1170 value
[symind
] = val
+1;
1172 } else if (*val
== '\0')
1173 value
[symind
] = "1";
1179 value
[symind
] = NULL
;
1181 debug("addsym %s=%s", symname
[symind
],
1182 value
[symind
] ? value
[symind
] : "undef");
1186 * Compare s with n characters of t.
1187 * The same as strncmp() except that it checks that s[n] == '\0'.
1190 strlcmp(const char *s
, const char *t
, size_t n
)
1192 while (n
-- && *t
!= '\0')
1194 return ((unsigned char)*s
- (unsigned char)*t
);
1197 return ((unsigned char)*s
);
1204 debug(const char *msg
, ...)
1216 error(const char *msg
)
1219 warnx("%s: %d: %s", filename
, linenum
, msg
);
1221 warnx("%s: %d: %s (#if line %d depth %d)",
1222 filename
, linenum
, msg
, stifline
[depth
], depth
);
1224 errx(2, "output may be truncated");