From: sthen Date: Fri, 18 Apr 2014 21:11:34 +0000 (+0000) Subject: Since we've been making heavy use of unifdef recently: update it to the X-Git-Url: http://artulab.com/gitweb/?a=commitdiff_plain;h=ebe1fdbdac3e636a07aeb99956e749e4a1cdc3d2;p=openbsd Since we've been making heavy use of unifdef recently: update it to the recent 2.10 release. "This code was derived from software contributed to Berkeley by Dave Yost. It was rewritten to support ANSI C by Tony Finch. The original version of unifdef carried the 4-clause BSD copyright licence. None of its code remains in this version (though some of the names remain) so it now carries a more liberal licence." ok deraadt@ --- diff --git a/usr.bin/unifdef/Makefile b/usr.bin/unifdef/Makefile index 9e47e467b2f..b9e41912780 100644 --- a/usr.bin/unifdef/Makefile +++ b/usr.bin/unifdef/Makefile @@ -1,6 +1,8 @@ -# $OpenBSD: Makefile,v 1.6 2003/11/20 23:23:09 avsm Exp $ +# $OpenBSD: Makefile,v 1.7 2014/04/18 21:11:34 sthen Exp $ PROG= unifdef COPTS+= -Wall +unifdef: unifdef.c unifdef.h version.h + .include diff --git a/usr.bin/unifdef/unifdef.1 b/usr.bin/unifdef/unifdef.1 index 0d8050cffa3..6bb86ef00b8 100644 --- a/usr.bin/unifdef/unifdef.1 +++ b/usr.bin/unifdef/unifdef.1 @@ -1,9 +1,9 @@ -.\" $OpenBSD: unifdef.1,v 1.18 2013/01/17 21:29:15 jmc Exp $ .\" Copyright (c) 1985, 1991, 1993 .\" The Regents of the University of California. All rights reserved. +.\" Copyright (c) 2002 - 2013 Tony Finch . All rights reserved. .\" .\" This code is derived from software contributed to Berkeley by -.\" Dave Yost. Support for #if and #elif was added by Tony Finch. +.\" Dave Yost. It was rewritten to support ANSI C by Tony Finch. .\" .\" Redistribution and use in source and binary forms, with or without .\" modification, are permitted provided that the following conditions @@ -29,11 +29,7 @@ .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF .\" SUCH DAMAGE. .\" -.\" @(#)unifdef.1 8.2 (Berkeley) 4/1/94 -.\" $dotat: things/unifdef.1,v 1.26 2002/09/24 19:44:12 fanf2 Exp $ -.\" $FreeBSD: src/usr.bin/unifdef/unifdef.1,v 1.15 2002/09/24 19:48:39 fanf Exp $ -.\" -.Dd $Mdocdate: January 17 2013 $ +.Dd $Mdocdate: April 18 2014 $ .Dt UNIFDEF 1 .Os .Sh NAME @@ -41,21 +37,16 @@ .Nd remove preprocessor conditionals from code .Sh SYNOPSIS .Nm -.Op Fl ceklst -.Sm off -.Oo -.Fl D Ar sym -.Op = Ar val -.Oc -.Op Fl I Ar path -.Oo -.Fl iD Ar sym -.Op = Ar val -.Oc -.Op Fl iU Ar sym -.Op Fl U Ar sym -.Sm on -.Op Ar file +.Op Fl bBcdehKkmnsStV +.Op Fl I Ns Ar path +.Op Fl [i]D Ns Ar sym Ns Op = Ns Ar val +.Op Fl [i]U Ns Ar sym +.Ar ... +.Op Fl f Ar defile +.Op Fl x Bro Ar 012 Brc +.Op Fl M Ar backext +.Op Fl o Ar outfile +.Op Ar infile ... .Sh DESCRIPTION The .Nm @@ -75,125 +66,216 @@ utility acts on and .Ic #endif lines, -and it understands only the commonly-used subset +using macros specified in +.Fl D +and +.Fl U +command line options or in +.Fl f +definitions files. +A directive is processed +if the macro specifications are sufficient to provide +a definite value for its control expression. +If the result is false, +the directive and the following lines under its control are removed. +If the result is true, +only the directive is removed. +An +.Ic #ifdef +or +.Ic #ifndef +directive is passed through unchanged +if its controlling macro is not specified. +Any +.Ic #if +or +.Ic #elif +control expression that has an unknown value or that +.Nm +cannot parse is passed through unchanged. +By default, +.Nm +ignores +.Ic #if +and +.Ic #elif +lines with constant expressions; +it can be told to process them by specifying the +.Fl k +flag on the command line. +.Pp +It understands a commonly-used subset of the expression syntax for .Ic #if and .Ic #elif -lines. -It handles -integer values of symbols defined on the command line, +lines: +integer constants, +integer values of macros defined on the command line, the .Fn defined -operator applied to symbols defined or undefined on the command line, +operator, the operators -.Ic \&! , < , > , <= , -.Ic >= , == , != , && , -.Ic || , +.Ic \&! , < , > , +.Ic <= , >= , == , != , +.Ic && , || , and parenthesized expressions. -Anything that it does not understand is passed through unharmed. -It only processes -.Ic #ifdef -and -.Ic #ifndef -directives if the symbol is specified on the command line, -otherwise they are also passed through unchanged. -By default, it ignores +A kind of +.Dq "short circuit" +evaluation is used for the +.Ic && +operator: +if either operand is definitely false then the result is false, +even if the value of the other operand is unknown. +Similarly, +if either operand of +.Ic || +is definitely true then the result is true. +.Pp +When evaluating an expression, +.Nm +does not expand macros first. +The value of a macro must be a simple number, +not an expression. +A limited form of indirection is allowed, +where one macro's value is the name of another. +.Pp +In most cases, +.Nm +does not distinguish between object-like macros +(without arguments) and function-like macros (with arguments). +A function-like macro invocation can appear in .Ic #if and .Ic #elif -lines with constant expressions, -or they may be processed by specifying the -.Fl k -flag on the command line. +control expressions. +If the macro is not explicitly defined, +or is defined with the +.Fl D +flag on the command-line, +or with +.Ic #define +in a +.Fl f +definitions file, +its arguments are ignored. +If a macro is explicitly undefined on the command line with the +.Fl U +flag, +or with +.Ic #undef +in a +.Fl f +definitions file, +it may not have any arguments since this leads to a syntax error. .Pp The .Nm -utility also understands just enough about C +utility understands just enough about C to know when one of the directives is inactive because it is inside a comment, or affected by a backslash-continued line. It spots unusually-formatted preprocessor directives -and knows when the layout is too odd to handle. +and knows when the layout is too odd for it to handle. +.Sh OPTIONS +.Bl -tag -width indent -compact +.It Fl D Ns Ar sym Ns = Ns Ar val +Specify that a macro is defined to a given value. .Pp -The options are as follows: +.It Fl D Ns Ar sym +Specify that a macro is defined to the value 1. +.Pp +.It Fl U Ns Ar sym +Specify that a macro is undefined. +.Pp +If the same macro appears in more than one argument, +the last occurrence dominates. +.Pp +.It Fl iD Ns Ar sym Ns Op = Ns Ar val +.It Fl iU Ns Ar sym +C strings, comments, +and line continuations +are ignored within +.Ic #ifdef +and +.Ic #ifndef +blocks +controlled by macros +specified with these options. +.Pp +.It Fl f Ar defile +The file +.Ar defile +contains +.Ic #define +and +.Ic #undef +preprocessor directives, +which have the same effect as the corresponding +.Fl D +and +.Fl U +command-line arguments. +You can have multiple +.Fl f +arguments and mix them with +.Fl D +and +.Fl U +arguments; +later options override earlier ones. +.Pp +Each directive must be on a single line. +Object-like macro definitions (without arguments) +are set to the given value. +Function-like macro definitions (with arguments) +are treated as if they are set to 1. +.Pp +.It Fl b +Replace removed lines with blank lines +instead of deleting them. +Mutually exclusive with the +.Fl B +option. +.Pp +.It Fl B +Compress blank lines around a deleted section. +Mutually exclusive with the +.Fl b +option. .Pp -.Bl -tag -width indent -compact .It Fl c -If the -.Fl c -flag is specified, -then the operation of -.Nm -is complemented, -i.e., the lines that would have been removed or blanked +Complement, +i.e., lines that would have been removed or blanked are retained and vice versa. .Pp -.Sm off -.It Xo -.Fl D Ar sym -.Op = Ar val -.Xc -.Sm on -Specify that a symbol is defined, -and optionally specify what value to give it -for the purpose of handling -.Ic #if -and -.Ic #elif -directives. +.It Fl d +Turn on printing of debugging messages. .Pp .It Fl e -Because -.Nm -processes its input one line at a time, -it cannot remove preprocessor directives that span more than one line. -The most common example of this is a directive with a multi-line -comment hanging off its right hand end. By default, -if .Nm -has to process such a directive, -it will complain that the line is too obfuscated. +will report an error if it needs to remove +a preprocessor directive that spans more than one line, +for example, if it has a multi-line +comment hanging off its right hand end. The .Fl e -option changes the behavior so that, -where possible, -such lines are left unprocessed instead of reporting an error. -.Pp -.Sm off -.It Xo -.Fl iD Ar sym -.Op = Ar val -.Xc -.Sm on -.It Fl iU Ns Ar sym -Ignore -.Ic #ifdef Ns s . -If your C code uses -.Ic #ifdef Ns s -to delimit non-C lines, -such as comments -or code which is under construction, -then you must tell -.Nm -which symbols are used for that purpose so that it will not try to parse -comments and line continuations -inside those -.Ic #ifdef Ns s . -One specifies ignored symbols with -.Sm off -.Fl iD Ar sym Op = Ar val -.Sm on -and -.Fl iU Ns Ar sym , -similar to -.Sm off -.Fl D Ar sym Op = Ar val -.Sm on +flag makes it ignore the line instead. +.Pp +.It Fl h +Print help. +.Pp +.It Fl K +Always treat the result of +.Ic && and -.Fl U Ns Ar sym . +.Ic || +operators as unknown if either operand is unknown, +instead of short-circuiting when unknown operands can't affect the result. +This option is for compatibility with older versions of +.Nm . .Pp .It Fl k Process @@ -203,51 +285,80 @@ and lines with constant expressions. By default, sections controlled by such lines are passed through unchanged because they typically start -.Li #if 0 +.Dq Li "#if 0" and are used as a kind of comment to sketch out future or past development. It would be rude to strip them out, just as it would be for normal comments. .Pp -.It Fl l -Replace removed lines with blank lines -instead of deleting them. +.It Fl m +Modify one or more input files in place. +.Pp +.It Fl M Ar backext +Modify input files in place, and keep backups of the original files by +appending the +.Ar backext +to the input filenames. +.Pp +.It Fl n +Add +.Li #line +directives to the output following any deleted lines, +so that errors produced when compiling the output file correspond to +line numbers in the input file. +.Pp +.It Fl o Ar outfile +Write output to the file +.Ar outfile +instead of the standard output when processing a single file. .Pp .It Fl s -Instead of processing the input file as usual, +Instead of processing an input file as usual, this option causes .Nm -to produce a list of symbols that appear in expressions -that -.Nm -understands. -It is useful in conjunction with the -.Fl dM -option of -.Xr cpp 1 -for creating -.Nm -command lines. +to produce a list of macros that are used in +preprocessor directive controlling expressions. +.Pp +.It Fl S +Like the +.Fl s +option, but the nesting depth of each macro is also printed. +This is useful for working out the number of possible combinations +of interdependent defined/undefined macros. .Pp .It Fl t -Disables parsing for C comments +Disables parsing for C strings, comments, and line continuations, which is useful for plain text. +This is a blanket version of the +.Fl iD +and +.Fl iU +flags. .Pp -.It Fl U Ns Ar sym -Specify that a symbol is undefined. -If the same symbol appears in more than one argument, -the last occurrence dominates. +.It Fl V +Print version details. +.Pp +.It Fl x Bro Ar 012 Brc +Set exit status mode to zero, one, or two. +See the +.Sx EXIT STATUS +section below for details. .El .Pp The .Nm -utility copies its output to -.Em stdout -and will take its input from +utility takes its input from .Em stdin -if no +if there are no .Ar file -argument is given. +arguments. +You must use the +.Fl m +or +.Fl M +options if there are multiple input files. +You can specify inut from stdin or output to stdout with +.Ql - . .Pp The .Nm @@ -256,10 +367,35 @@ utility works nicely with the option of .Xr diff 1 . .Sh EXIT STATUS -The +In normal usage the .Nm -utility exits 0 if the output is an exact copy of the input, -1 if not, and 2 if in trouble. +utility's exit status depends on the mode set using the +.Fl x +option. +.Pp +If the exit mode is zero (the default) then +.Nm +exits with status 0 if the output is an exact copy of the input, +or with status 1 if the output differs. +.Pp +If the exit mode is one, +.Nm +exits with status 1 if the output is unmodified +or 0 if it differs. +.Pp +If the exit mode is two, +.Nm +exits with status zero in both cases. +.Pp +In all exit modes, +.Nm +exits with status 2 if there is an error. +.Pp +The exit status is 0 if the +.Fl h +or +.Fl V +command line options are given. .Sh DIAGNOSTICS .Bl -item .It @@ -284,23 +420,36 @@ in comment. .Sh SEE ALSO .Xr cpp 1 , .Xr diff 1 +.Pp +The unifdef home page is +.Pa http://dotat.at/prog/unifdef .Sh HISTORY The .Nm command appeared in -.Bx 4.3 . +.Bx 2.9 . +.Tn ANSI\~C +support was added in +.Fx 4.7 . +.Sh AUTHORS +The original implementation was written by +.An Dave Yost Aq Dave@Yost.com . +.An Tony Finch Aq dot@dotat.at +rewrote it to support +.Tn ANSI\~C . .Sh BUGS Expression evaluation is very limited. .Pp -Preprocessor control lines split across more than one physical line +Handling one line at a time means +preprocessor directives split across more than one physical line (because of comments or backslash-newline) cannot be handled in every situation. .Pp Trigraphs are not recognized. .Pp -There is no support for symbols with different definitions at +There is no support for macros with different definitions at different points in the source file. .Pp -The text-mode and ignore functionality doesn't correspond to modern +The text-mode and ignore functionality does not correspond to modern .Xr cpp 1 behaviour. diff --git a/usr.bin/unifdef/unifdef.c b/usr.bin/unifdef/unifdef.c index c96f11f7eea..65ddb1e04be 100644 --- a/usr.bin/unifdef/unifdef.c +++ b/usr.bin/unifdef/unifdef.c @@ -1,11 +1,5 @@ -/* $OpenBSD: unifdef.c,v 1.15 2012/03/04 04:05:15 fgsch Exp $ */ /* - * Copyright (c) 2002, 2003 Tony Finch - * Copyright (c) 1985, 1993 - * The Regents of the University of California. All rights reserved. - * - * This code is derived from software contributed to Berkeley by - * Dave Yost. Support for #if and #elif was added by Tony Finch. + * Copyright (c) 2002 - 2014 Tony Finch * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions @@ -15,14 +9,11 @@ * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. - * 3. Neither the name of the University nor the names of its contributors - * may be used to endorse or promote products derived from this software - * without specific prior written permission. * - * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE - * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) @@ -35,26 +26,30 @@ /* * unifdef - remove ifdef'ed lines * + * This code was derived from software contributed to Berkeley by Dave Yost. + * It was rewritten to support ANSI C by Tony Finch. The original version + * of unifdef carried the 4-clause BSD copyright licence. None of its code + * remains in this version (though some of the names remain) so it now + * carries a more liberal licence. + * * Wishlist: * provide an option which will append the name of the * appropriate symbol after #else's and #endif's * provide an option which will check symbols after * #else's and #endif's to see that they match their * corresponding #ifdef or #ifndef - * generate #line directives in place of deleted code * - * The first two items above require better buffer handling, which would - * also make it possible to handle all "dodgy" directives correctly. + * These require better buffer handling, which would also make + * it possible to handle all "dodgy" directives correctly. */ -#include -#include -#include -#include -#include -#include -#include -#include +#include "unifdef.h" + +static const char copyright[] = + #include "version.h" + "@(#) $Author: sthen $\n" + "@(#) $URL: http://dotat.at/prog/unifdef $\n" +; /* types of input lines: */ typedef enum { @@ -72,6 +67,7 @@ typedef enum { LT_DODGY_LAST = LT_DODGY + LT_ENDIF, LT_PLAIN, /* ordinary line */ LT_EOF, /* end of file */ + LT_ERROR, /* unevaluable #if */ LT_COUNT } Linetype; @@ -82,9 +78,12 @@ static char const * const linetype_name[] = { "DODGY IF", "DODGY TRUE", "DODGY FALSE", "DODGY ELIF", "DODGY ELTRUE", "DODGY ELFALSE", "DODGY ELSE", "DODGY ENDIF", - "PLAIN", "EOF" + "PLAIN", "EOF", "ERROR" }; +#define linetype_if2elif(lt) ((Linetype)(lt - LT_IF + LT_ELIF)) +#define linetype_2dodgy(lt) ((Linetype)(lt + LT_DODGY)) + /* state of #if processing */ typedef enum { IS_OUTSIDE, @@ -113,11 +112,13 @@ typedef enum { C_COMMENT, /* in a comment like this one */ CXX_COMMENT, /* between // and end of line */ STARTING_COMMENT, /* just after slash-backslash-newline */ - FINISHING_COMMENT /* star-backslash-newline in a C comment */ + FINISHING_COMMENT, /* star-backslash-newline in a C comment */ + CHAR_LITERAL, /* inside '' */ + STRING_LITERAL /* inside "" */ } Comment_state; static char const * const comment_name[] = { - "NO", "C", "CXX", "STARTING", "FINISHING" + "NO", "C", "CXX", "STARTING", "FINISHING", "CHAR", "STRING" }; /* state of preprocessor line parser */ @@ -136,7 +137,7 @@ static char const * const linestate_name[] = { */ #define MAXDEPTH 64 /* maximum #if nesting */ #define MAXLINE 4096 /* maximum length of line */ -#define MAXSYMS 4096 /* maximum number of symbols */ +#define MAXSYMS 16384 /* maximum number of symbols */ /* * Sometimes when editing a keyword the replacement text is longer, so @@ -148,12 +149,17 @@ static char const * const linestate_name[] = { * Globals. */ +static bool compblank; /* -B: compress blank lines */ +static bool lnblank; /* -b: blank deleted lines */ static bool complement; /* -c: do the complement */ static bool debugging; /* -d: debugging reports */ +static bool inplace; /* -m: modify in place */ static bool iocccok; /* -e: fewer IOCCC errors */ +static bool strictlogic; /* -K: keep ambiguous #ifs */ static bool killconsts; /* -k: eval constant #ifs */ -static bool lnblank; /* -l: blank deleted lines */ +static bool lnnum; /* -n: add #line directives */ static bool symlist; /* -s: output symbol list */ +static bool symdepth; /* -S: output symbol depth */ static bool text; /* -t: this is a text file */ static const char *symname[MAXSYMS]; /* symbol name */ @@ -164,39 +170,81 @@ static int nsyms; /* number of symbols */ static FILE *input; /* input file pointer */ static const char *filename; /* input file name */ static int linenum; /* current line number */ +static const char *linefile; /* file name for #line */ +static FILE *output; /* output file pointer */ +static const char *ofilename; /* output file name */ +static const char *backext; /* backup extension */ +static char *tempname; /* avoid splatting input */ static char tline[MAXLINE+EDITSLOP];/* input buffer plus space */ static char *keyword; /* used for editing #elif's */ +/* + * When processing a file, the output's newline style will match the + * input's, and unifdef correctly handles CRLF or LF endings whatever + * the platform's native style. The stdio streams are opened in binary + * mode to accommodate platforms whose native newline style is CRLF. + * When the output isn't a processed input file (when it is error / + * debug / diagnostic messages) then unifdef uses native line endings. + */ + +static const char *newline; /* input file format */ +static const char newline_unix[] = "\n"; +static const char newline_crlf[] = "\r\n"; + static Comment_state incomment; /* comment parser state */ static Line_state linestate; /* #if line parser state */ static Ifstate ifstate[MAXDEPTH]; /* #if processor state */ static bool ignoring[MAXDEPTH]; /* ignore comments state */ static int stifline[MAXDEPTH]; /* start of current #if */ static int depth; /* current #if nesting */ -static bool keepthis; /* don't delete constant #if */ - +static int delcount; /* count of deleted lines */ +static unsigned blankcount; /* count of blank lines */ +static unsigned blankmax; /* maximum recent blankcount */ +static bool constexpr; /* constant #if expression */ +static bool zerosyms; /* to format symdepth output */ +static bool firstsym; /* ditto */ + +static int exitmode; /* exit status mode */ static int exitstat; /* program exit status */ -static void addsym(bool, bool, char *); +static void addsym1(bool, bool, char *); +static void addsym2(bool, const char *, const char *); +static char *astrcat(const char *, const char *); +static void cleantemp(void); +static void closeio(void); static void debug(const char *, ...); +static void debugsym(const char *, int); +static bool defundef(void); +static void defundefile(const char *); +static void done(void); static void error(const char *); -static int findsym(const char *); +static int findsym(const char **); static void flushline(bool); -static Linetype get_line(void); +static void hashline(void); +static void help(void); static Linetype ifeval(const char **); static void ignoreoff(void); static void ignoreon(void); +static void indirectsym(void); static void keywordedit(const char *); +static const char *matchsym(const char *, const char *); static void nest(void); +static Linetype parseline(void); static void process(void); +static void processinout(const char *, const char *); +static const char *skipargs(const char *); static const char *skipcomment(const char *); +static const char *skiphash(void); +static const char *skipline(const char *); static const char *skipsym(const char *); static void state(Ifstate); -static int strlcmp(const char *, const char *, size_t); +static void unnest(void); static void usage(void); +static void version(void); +static const char *xstrdup(const char *, const char *); -#define endsym(c) (!isalpha((unsigned char)c) && !isdigit((unsigned char)c) && c != '_') +#define endsym(c) (!isalnum((unsigned char)c) && c != '_') /* * The main program. @@ -206,7 +254,7 @@ main(int argc, char *argv[]) { int opt; - while ((opt = getopt(argc, argv, "i:D:U:I:cdeklst")) != -1) + while ((opt = getopt(argc, argv, "i:D:U:f:I:M:o:x:bBcdehKklmnsStV")) != -1) switch (opt) { case 'i': /* treat stuff controlled by these symbols as text */ /* @@ -216,20 +264,26 @@ main(int argc, char *argv[]) */ opt = *optarg++; if (opt == 'D') - addsym(true, true, optarg); + addsym1(true, true, optarg); else if (opt == 'U') - addsym(true, false, optarg); + addsym1(true, false, optarg); else usage(); break; case 'D': /* define a symbol */ - addsym(false, true, optarg); + addsym1(false, true, optarg); break; case 'U': /* undef a symbol */ - addsym(false, false, optarg); + addsym1(false, false, optarg); break; - case 'I': - /* no-op for compatibility with cpp */ + case 'I': /* no-op for compatibility with cpp */ + break; + case 'b': /* blank deleted lines instead of omitting them */ + case 'l': /* backwards compatibility */ + lnblank = true; + break; + case 'B': /* compress blank lines around removed section */ + compblank = true; break; case 'c': /* treat -D as -U and vice versa */ complement = true; @@ -240,60 +294,214 @@ main(int argc, char *argv[]) case 'e': /* fewer errors from dodgy lines */ iocccok = true; break; + case 'f': /* definitions file */ + defundefile(optarg); + break; + case 'h': + help(); + break; + case 'K': /* keep ambiguous #ifs */ + strictlogic = true; + break; case 'k': /* process constant #ifs */ killconsts = true; break; - case 'l': /* blank deleted lines instead of omitting them */ - lnblank = true; + case 'm': /* modify in place */ + inplace = true; + break; + case 'M': /* modify in place and keep backup */ + inplace = true; + backext = optarg; + break; + case 'n': /* add #line directive after deleted lines */ + lnnum = true; + break; + case 'o': /* output to a file */ + ofilename = optarg; break; case 's': /* only output list of symbols that control #ifs */ symlist = true; break; + case 'S': /* list symbols with their nesting depth */ + symlist = symdepth = true; + break; case 't': /* don't parse C comments */ text = true; break; + case 'V': + version(); + break; + case 'x': + exitmode = atoi(optarg); + if(exitmode < 0 || exitmode > 2) + usage(); + break; default: usage(); } argc -= optind; argv += optind; - if (nsyms == 0 && !symlist) { - warnx("must -D or -U at least one symbol"); - usage(); + if (compblank && lnblank) + errx(2, "-B and -b are mutually exclusive"); + if (symlist && (ofilename != NULL || inplace || argc > 1)) + errx(2, "-s only works with one input file"); + if (argc > 1 && ofilename != NULL) + errx(2, "-o cannot be used with multiple input files"); + if (argc > 1 && !inplace) + errx(2, "multiple input files require -m or -M"); + if (argc == 0) + argc = 1; + if (argc == 1 && !inplace && ofilename == NULL) + ofilename = "-"; + indirectsym(); + + atexit(cleantemp); + if (ofilename != NULL) + processinout(*argv, ofilename); + else while (argc-- > 0) { + processinout(*argv, *argv); + argv++; } - if (argc > 1) { - errx(2, "can only do one file"); - } else if (argc == 1 && strcmp(*argv, "-") != 0) { - filename = *argv; - if ((input = fopen(filename, "r")) != NULL) { - process(); - (void) fclose(input); - } else - err(2, "can't open %s", *argv); - } else { + switch(exitmode) { + case(0): exit(exitstat); + case(1): exit(!exitstat); + case(2): exit(0); + default: abort(); /* bug */ + } +} + +/* + * File logistics. + */ +static void +processinout(const char *ifn, const char *ofn) +{ + struct stat st; + + if (ifn == NULL || strcmp(ifn, "-") == 0) { filename = "[stdin]"; - input = stdin; + linefile = NULL; + input = fbinmode(stdin); + } else { + filename = ifn; + linefile = ifn; + input = fopen(ifn, "rb"); + if (input == NULL) + err(2, "can't open %s", ifn); + } + if (strcmp(ofn, "-") == 0) { + output = fbinmode(stdout); + process(); + return; + } + if (stat(ofn, &st) < 0) { + output = fopen(ofn, "wb"); + if (output == NULL) + err(2, "can't create %s", ofn); process(); + return; + } + + tempname = astrcat(ofn, ".XXXXXX"); + output = mktempmode(tempname, st.st_mode); + if (output == NULL) + err(2, "can't create %s", tempname); + + process(); + + if (backext != NULL) { + char *backname = astrcat(ofn, backext); + if (rename(ofn, backname) < 0) + err(2, "can't rename \"%s\" to \"%s\"", ofn, backname); + free(backname); } + if (replace(tempname, ofn) < 0) + err(2, "can't rename \"%s\" to \"%s\"", tempname, ofn); + free(tempname); + tempname = NULL; +} - exit(exitstat); +/* + * For cleaning up if there is an error. + */ +static void +cleantemp(void) +{ + if (tempname != NULL) + remove(tempname); +} + +/* + * Self-identification functions. + */ + +static void +version(void) +{ + const char *c = copyright; + for (;;) { + while (*++c != '$') + if (*c == '\0') + exit(0); + while (*++c != '$') + putc(*c, stderr); + putc('\n', stderr); + } +} + +static void +synopsis(FILE *fp) +{ + fprintf(fp, + "usage: unifdef [-bBcdehKkmnsStV] [-x{012}] [-Mext] [-opath] \\\n" + " [-[i]Dsym[=val]] [-[i]Usym] [-fpath] ... [file] ...\n"); } static void usage(void) { - fprintf(stderr, - "usage: unifdef [-ceklst] [-Dsym[=val]] [-Ipath] [-iDsym[=val]] " - "[-iUsym] [-Usym]\n" - "\t[file]\n"); + synopsis(stderr); exit(2); } +static void +help(void) +{ + synopsis(stdout); + printf( + " -Dsym=val define preprocessor symbol with given value\n" + " -Dsym define preprocessor symbol with value 1\n" + " -Usym preprocessor symbol is undefined\n" + " -iDsym=val \\ ignore C strings and comments\n" + " -iDsym ) in sections controlled by these\n" + " -iUsym / preprocessor symbols\n" + " -fpath file containing #define and #undef directives\n" + " -b blank lines instead of deleting them\n" + " -B compress blank lines around deleted section\n" + " -c complement (invert) keep vs. delete\n" + " -d debugging mode\n" + " -e ignore multiline preprocessor directives\n" + " -h print help\n" + " -Ipath extra include file path (ignored)\n" + " -K disable && and || short-circuiting\n" + " -k process constant #if expressions\n" + " -Mext modify in place and keep backups\n" + " -m modify input files in place\n" + " -n add #line directives to output\n" + " -opath output file name\n" + " -S list #if control symbols with nesting\n" + " -s list #if control symbols\n" + " -t ignore C strings and comments\n" + " -V print version\n" + " -x{012} exit status mode\n" + ); + exit(0); +} + /* * A state transition function alters the global #if processing state * in a particular way. The table below is indexed by the current - * processing state and the type of the current line. A NULL entry - * indicates that processing is complete. + * processing state and the type of the current line. * * Nesting is handled by keeping a stack of states; some transition * functions increase or decrease the depth. They also maintain the @@ -303,7 +511,8 @@ usage(void) * When we have processed a group that starts off with a known-false * #if/#elif sequence (which has therefore been deleted) followed by a * #elif that we don't understand and therefore must keep, we edit the - * latter into a #if to keep the nesting correct. + * latter into a #if to keep the nesting correct. We use memcpy() to + * overwrite the 4 byte token "elif" with "if " without a '\0' byte. * * When we find a true #elif in a group, the following block will * always be kept and the rest of the sequence after the next #elif or @@ -322,280 +531,90 @@ usage(void) typedef void state_fn(void); /* report an error */ -static void -Eelif(void) -{ - error("Inappropriate #elif"); -} - -static void -Eelse(void) -{ - error("Inappropriate #else"); -} - -static void -Eendif(void) -{ - error("Inappropriate #endif"); -} - -static void -Eeof(void) -{ - error("Premature EOF"); -} - -static void -Eioccc(void) -{ - error("Obfuscated preprocessor control line"); -} - +static void Eelif (void) { error("Inappropriate #elif"); } +static void Eelse (void) { error("Inappropriate #else"); } +static void Eendif(void) { error("Inappropriate #endif"); } +static void Eeof (void) { error("Premature EOF"); } +static void Eioccc(void) { error("Obfuscated preprocessor control line"); } /* plain line handling */ -static void -print(void) -{ - flushline(true); -} - -static void -drop(void) -{ - flushline(false); -} - +static void print (void) { flushline(true); } +static void drop (void) { flushline(false); } /* output lacks group's start line */ -static void -Strue(void) -{ - drop(); - ignoreoff(); - state(IS_TRUE_PREFIX); -} - -static void -Sfalse(void) -{ - drop(); - ignoreoff(); - state(IS_FALSE_PREFIX); -} - -static void -Selse(void) -{ - drop(); - state(IS_TRUE_ELSE); -} - +static void Strue (void) { drop(); ignoreoff(); state(IS_TRUE_PREFIX); } +static void Sfalse(void) { drop(); ignoreoff(); state(IS_FALSE_PREFIX); } +static void Selse (void) { drop(); state(IS_TRUE_ELSE); } /* print/pass this block */ -static void -Pelif(void) -{ - print(); - ignoreoff(); - state(IS_PASS_MIDDLE); -} - -static void -Pelse(void) -{ - print(); - state(IS_PASS_ELSE); -} - -static void -Pendif(void) -{ - print(); - --depth; -} - +static void Pelif (void) { print(); ignoreoff(); state(IS_PASS_MIDDLE); } +static void Pelse (void) { print(); state(IS_PASS_ELSE); } +static void Pendif(void) { print(); unnest(); } /* discard this block */ -static void -Dfalse(void) -{ - drop(); - ignoreoff(); - state(IS_FALSE_TRAILER); -} - -static void -Delif(void) -{ - drop(); - ignoreoff(); - state(IS_FALSE_MIDDLE); -} - -static void -Delse(void) -{ - drop(); - state(IS_FALSE_ELSE); -} - -static void -Dendif(void) -{ - drop(); - --depth; -} - +static void Dfalse(void) { drop(); ignoreoff(); state(IS_FALSE_TRAILER); } +static void Delif (void) { drop(); ignoreoff(); state(IS_FALSE_MIDDLE); } +static void Delse (void) { drop(); state(IS_FALSE_ELSE); } +static void Dendif(void) { drop(); unnest(); } /* first line of group */ -static void -Fdrop(void) -{ - nest(); - Dfalse(); -} - -static void -Fpass(void) -{ - nest(); - Pelif(); -} - -static void -Ftrue(void) -{ - nest(); - Strue(); -} - -static void -Ffalse(void) -{ - nest(); - Sfalse(); -} - +static void Fdrop (void) { nest(); Dfalse(); } +static void Fpass (void) { nest(); Pelif(); } +static void Ftrue (void) { nest(); Strue(); } +static void Ffalse(void) { nest(); Sfalse(); } /* variable pedantry for obfuscated lines */ -static void -Oiffy(void) -{ - if (iocccok) - Fpass(); - else - Eioccc(); - ignoreon(); -} - -static void -Oif(void) -{ - if (iocccok) - Fpass(); - else - Eioccc(); -} - -static void -Oelif(void) -{ - if (iocccok) - Pelif(); - else - Eioccc(); -} - +static void Oiffy (void) { if (!iocccok) Eioccc(); Fpass(); ignoreon(); } +static void Oif (void) { if (!iocccok) Eioccc(); Fpass(); } +static void Oelif (void) { if (!iocccok) Eioccc(); Pelif(); } /* ignore comments in this block */ -static void -Idrop(void) -{ - Fdrop(); - ignoreon(); -} - -static void -Itrue(void) -{ - Ftrue(); - ignoreon(); -} - -static void -Ifalse(void) -{ - Ffalse(); - ignoreon(); -} - -/* edit this line */ -static void -Mpass (void) -{ - strncpy(keyword, "if ", 4); - Pelif(); -} - -static void -Mtrue (void) -{ - keywordedit("else\n"); - state(IS_TRUE_MIDDLE); -} - -static void -Melif (void) -{ - keywordedit("endif\n"); - state(IS_FALSE_TRAILER); -} - -static void -Melse (void) -{ - keywordedit("endif\n"); - state(IS_FALSE_ELSE); -} +static void Idrop (void) { Fdrop(); ignoreon(); } +static void Itrue (void) { Ftrue(); ignoreon(); } +static void Ifalse(void) { Ffalse(); ignoreon(); } +/* modify this line */ +static void Mpass (void) { memcpy(keyword, "if ", 4); Pelif(); } +static void Mtrue (void) { keywordedit("else"); state(IS_TRUE_MIDDLE); } +static void Melif (void) { keywordedit("endif"); state(IS_FALSE_TRAILER); } +static void Melse (void) { keywordedit("endif"); state(IS_FALSE_ELSE); } static state_fn * const trans_table[IS_COUNT][LT_COUNT] = { /* IS_OUTSIDE */ { Itrue, Ifalse,Fpass, Ftrue, Ffalse,Eelif, Eelif, Eelif, Eelse, Eendif, Oiffy, Oiffy, Fpass, Oif, Oif, Eelif, Eelif, Eelif, Eelse, Eendif, - print, NULL }, + print, done, abort }, /* IS_FALSE_PREFIX */ { Idrop, Idrop, Fdrop, Fdrop, Fdrop, Mpass, Strue, Sfalse,Selse, Dendif, Idrop, Idrop, Fdrop, Fdrop, Fdrop, Mpass, Eioccc,Eioccc,Eioccc,Eioccc, - drop, Eeof }, + drop, Eeof, abort }, /* IS_TRUE_PREFIX */ { Itrue, Ifalse,Fpass, Ftrue, Ffalse,Dfalse,Dfalse,Dfalse,Delse, Dendif, Oiffy, Oiffy, Fpass, Oif, Oif, Eioccc,Eioccc,Eioccc,Eioccc,Eioccc, - print, Eeof }, + print, Eeof, abort }, /* IS_PASS_MIDDLE */ { Itrue, Ifalse,Fpass, Ftrue, Ffalse,Pelif, Mtrue, Delif, Pelse, Pendif, Oiffy, Oiffy, Fpass, Oif, Oif, Pelif, Oelif, Oelif, Pelse, Pendif, - print, Eeof }, + print, Eeof, abort }, /* IS_FALSE_MIDDLE */ { Idrop, Idrop, Fdrop, Fdrop, Fdrop, Pelif, Mtrue, Delif, Pelse, Pendif, Idrop, Idrop, Fdrop, Fdrop, Fdrop, Eioccc,Eioccc,Eioccc,Eioccc,Eioccc, - drop, Eeof }, + drop, Eeof, abort }, /* IS_TRUE_MIDDLE */ { Itrue, Ifalse,Fpass, Ftrue, Ffalse,Melif, Melif, Melif, Melse, Pendif, Oiffy, Oiffy, Fpass, Oif, Oif, Eioccc,Eioccc,Eioccc,Eioccc,Pendif, - print, Eeof }, + print, Eeof, abort }, /* IS_PASS_ELSE */ { Itrue, Ifalse,Fpass, Ftrue, Ffalse,Eelif, Eelif, Eelif, Eelse, Pendif, Oiffy, Oiffy, Fpass, Oif, Oif, Eelif, Eelif, Eelif, Eelse, Pendif, - print, Eeof }, + print, Eeof, abort }, /* IS_FALSE_ELSE */ { Idrop, Idrop, Fdrop, Fdrop, Fdrop, Eelif, Eelif, Eelif, Eelse, Dendif, Idrop, Idrop, Fdrop, Fdrop, Fdrop, Eelif, Eelif, Eelif, Eelse, Eioccc, - drop, Eeof }, + drop, Eeof, abort }, /* IS_TRUE_ELSE */ { Itrue, Ifalse,Fpass, Ftrue, Ffalse,Eelif, Eelif, Eelif, Eelse, Dendif, Oiffy, Oiffy, Fpass, Oif, Oif, Eelif, Eelif, Eelif, Eelse, Eioccc, - print, Eeof }, + print, Eeof, abort }, /* IS_FALSE_TRAILER */ { Idrop, Idrop, Fdrop, Fdrop, Fdrop, Dfalse,Dfalse,Dfalse,Delse, Dendif, Idrop, Idrop, Fdrop, Fdrop, Fdrop, Dfalse,Dfalse,Dfalse,Delse, Eioccc, - drop, Eeof } + drop, Eeof, abort } /*TRUEI FALSEI IF TRUE FALSE ELIF ELTRUE ELFALSE ELSE ENDIF TRUEI FALSEI IF TRUE FALSE ELIF ELTRUE ELFALSE ELSE ENDIF (DODGY) - PLAIN EOF */ + PLAIN EOF ERROR */ }; /* @@ -604,52 +623,119 @@ static state_fn * const trans_table[IS_COUNT][LT_COUNT] = { static void ignoreoff(void) { + if (depth == 0) + abort(); /* bug */ ignoring[depth] = ignoring[depth-1]; } - static void ignoreon(void) { ignoring[depth] = true; } - static void keywordedit(const char *replacement) { - strlcpy(keyword, replacement, tline + sizeof(tline) - keyword); + snprintf(keyword, tline + sizeof(tline) - keyword, + "%s%s", replacement, newline); print(); } - static void nest(void) { - depth += 1; - if (depth >= MAXDEPTH) + if (depth > MAXDEPTH-1) + abort(); /* bug */ + if (depth == MAXDEPTH-1) error("Too many levels of nesting"); + depth += 1; stifline[depth] = linenum; } - +static void +unnest(void) +{ + if (depth == 0) + abort(); /* bug */ + depth -= 1; +} static void state(Ifstate is) { ifstate[depth] = is; } +/* + * The last state transition function. When this is called, + * lineval == LT_EOF, so the process() loop will terminate. + */ +static void +done(void) +{ + if (incomment) + error("EOF in comment"); + closeio(); +} + /* * Write a line to the output or not, according to command line options. + * If writing fails, closeio() will print the error and exit. */ static void flushline(bool keep) { if (symlist) return; - if (keep ^ complement) - fputs(tline, stdout); - else { - if (lnblank) - putc('\n', stdout); + if (keep ^ complement) { + bool blankline = tline[strspn(tline, " \t\r\n")] == '\0'; + if (blankline && compblank && blankcount != blankmax) { + delcount += 1; + blankcount += 1; + } else { + if (lnnum && delcount > 0) + hashline(); + if (fputs(tline, output) == EOF) + closeio(); + delcount = 0; + blankmax = blankcount = blankline ? blankcount + 1 : 0; + } + } else { + if (lnblank && fputs(newline, output) == EOF) + closeio(); exitstat = 1; + delcount += 1; + blankcount = 0; } + if (debugging && fflush(output) == EOF) + closeio(); +} + +/* + * Format of #line directives depends on whether we know the input filename. + */ +static void +hashline(void) +{ + int e; + + if (linefile == NULL) + e = fprintf(output, "#line %d%s", linenum, newline); + else + e = fprintf(output, "#line %d \"%s\"%s", + linenum, linefile, newline); + if (e < 0) + closeio(); +} + +/* + * Flush the output and handle errors. + */ +static void +closeio(void) +{ + /* Tidy up after findsym(). */ + if (symdepth && !zerosyms) + printf("\n"); + if (output != NULL && (ferror(output) || fclose(output) == EOF)) + err(2, "%s: can't write to output", filename); + fclose(input); } /* @@ -658,162 +744,152 @@ flushline(bool keep) static void process(void) { - Linetype lineval; - state_fn *trans; - - for (;;) { - linenum++; - lineval = get_line(); - trans = trans_table[ifstate[depth]][lineval]; - if (trans == NULL) - break; - trans(); - debug("process %s -> %s depth %d", - linetype_name[lineval], + Linetype lineval = LT_PLAIN; + /* When compressing blank lines, act as if the file + is preceded by a large number of blank lines. */ + blankmax = blankcount = 1000; + zerosyms = true; + newline = NULL; + linenum = 0; + while (lineval != LT_EOF) { + lineval = parseline(); + trans_table[ifstate[depth]][lineval](); + debug("process line %d %s -> %s depth %d", + linenum, linetype_name[lineval], ifstate_name[ifstate[depth]], depth); } - if (incomment) - error("EOF in comment"); } /* * Parse a line and determine its type. We keep the preprocessor line - * parser state between calls in a global variable. + * parser state between calls in the global variable linestate, with + * help from skipcomment(). */ static Linetype -get_line(void) +parseline(void) { const char *cp; int cursym; - int kwlen; Linetype retval; Comment_state wascomment; - if (fgets(tline, MAXLINE, input) == NULL) - return (LT_EOF); - retval = LT_PLAIN; wascomment = incomment; - cp = skipcomment(tline); - if (linestate == LS_START) { - if (*cp == '#') { - linestate = LS_HASH; - cp = skipcomment(cp + 1); - } else if (*cp != '\0') - linestate = LS_DIRTY; + cp = skiphash(); + if (cp == NULL) + return (LT_EOF); + if (newline == NULL) { + if (strrchr(tline, '\n') == strrchr(tline, '\r') + 1) + newline = newline_crlf; + else + newline = newline_unix; } - if (!incomment && linestate == LS_HASH) { - keyword = tline + (cp - tline); - cp = skipsym(cp); - kwlen = cp - keyword; - /* no way can we deal with a continuation inside a keyword */ - if (strncmp(cp, "\\\n", 2) == 0) - Eioccc(); - if (strlcmp("ifdef", keyword, kwlen) == 0 || - strlcmp("ifndef", keyword, kwlen) == 0) { - cp = skipcomment(cp); - if ((cursym = findsym(cp)) < 0) - retval = LT_IF; - else { - retval = (keyword[2] == 'n') - ? LT_FALSE : LT_TRUE; - if (value[cursym] == NULL) - retval = (retval == LT_TRUE) - ? LT_FALSE : LT_TRUE; - if (ignore[cursym]) - retval = (retval == LT_TRUE) - ? LT_TRUEI : LT_FALSEI; - } - cp = skipsym(cp); - } else if (strlcmp("if", keyword, kwlen) == 0) - retval = ifeval(&cp); - else if (strlcmp("elif", keyword, kwlen) == 0) - retval = ifeval(&cp) - LT_IF + LT_ELIF; - else if (strlcmp("else", keyword, kwlen) == 0) - retval = LT_ELSE; - else if (strlcmp("endif", keyword, kwlen) == 0) - retval = LT_ENDIF; + if (*cp == '\0') { + retval = LT_PLAIN; + goto done; + } + keyword = tline + (cp - tline); + if ((cp = matchsym("ifdef", keyword)) != NULL || + (cp = matchsym("ifndef", keyword)) != NULL) { + cp = skipcomment(cp); + if ((cursym = findsym(&cp)) < 0) + retval = LT_IF; else { - linestate = LS_DIRTY; - retval = LT_PLAIN; + retval = (keyword[2] == 'n') + ? LT_FALSE : LT_TRUE; + if (value[cursym] == NULL) + retval = (retval == LT_TRUE) + ? LT_FALSE : LT_TRUE; + if (ignore[cursym]) + retval = (retval == LT_TRUE) + ? LT_TRUEI : LT_FALSEI; } - cp = skipcomment(cp); - if (*cp != '\0') { + } else if ((cp = matchsym("if", keyword)) != NULL) + retval = ifeval(&cp); + else if ((cp = matchsym("elif", keyword)) != NULL) + retval = linetype_if2elif(ifeval(&cp)); + else if ((cp = matchsym("else", keyword)) != NULL) + retval = LT_ELSE; + else if ((cp = matchsym("endif", keyword)) != NULL) + retval = LT_ENDIF; + else { + cp = skipsym(keyword); + /* no way can we deal with a continuation inside a keyword */ + if (strncmp(cp, "\\\r\n", 3) == 0 || + strncmp(cp, "\\\n", 2) == 0) + Eioccc(); + cp = skipline(cp); + retval = LT_PLAIN; + goto done; + } + cp = skipcomment(cp); + if (*cp != '\0') { + cp = skipline(cp); + if (retval == LT_TRUE || retval == LT_FALSE || + retval == LT_TRUEI || retval == LT_FALSEI) + retval = LT_IF; + if (retval == LT_ELTRUE || retval == LT_ELFALSE) + retval = LT_ELIF; + } + /* the following can happen if the last line of the file lacks a + newline or if there is too much whitespace in a directive */ + if (linestate == LS_HASH) { + long len = cp - tline; + if (fgets(tline + len, MAXLINE - len, input) == NULL) { + if (ferror(input)) + err(2, "can't read %s", filename); + /* append the missing newline at eof */ + strcpy(tline + len, newline); + cp += strlen(newline); + linestate = LS_START; + } else { linestate = LS_DIRTY; - if (retval == LT_TRUE || retval == LT_FALSE || - retval == LT_TRUEI || retval == LT_FALSEI) - retval = LT_IF; - if (retval == LT_ELTRUE || retval == LT_ELFALSE) - retval = LT_ELIF; - } - if (retval != LT_PLAIN && (wascomment || incomment)) { - retval += LT_DODGY; - if (incomment) - linestate = LS_DIRTY; } - /* skipcomment should have changed the state */ - if (linestate == LS_HASH) - abort(); /* bug */ } - if (linestate == LS_DIRTY) { - while (*cp != '\0') - cp = skipcomment(cp + 1); + if (retval != LT_PLAIN && (wascomment || linestate != LS_START)) { + retval = linetype_2dodgy(retval); + linestate = LS_DIRTY; } - debug("parser %s comment %s line", +done: + debug("parser line %d state %s comment %s line", linenum, comment_name[incomment], linestate_name[linestate]); return (retval); } /* - * These are the operators that are supported by the expression - * evaluator. Note that if support for division is added then we also - * need short-circuiting booleans because of divide-by-zero. + * These are the binary operators that are supported by the expression + * evaluator. */ -static int -op_lt(int a, int b) -{ - return (a < b); +static Linetype op_strict(long *p, long v, Linetype at, Linetype bt) { + if(at == LT_IF || bt == LT_IF) return (LT_IF); + return (*p = v, v ? LT_TRUE : LT_FALSE); } - -static int -op_gt(int a, int b) -{ - return (a > b); +static Linetype op_lt(long *p, Linetype at, long a, Linetype bt, long b) { + return op_strict(p, a < b, at, bt); } - -static int -op_le(int a, int b) -{ - return (a <= b); +static Linetype op_gt(long *p, Linetype at, long a, Linetype bt, long b) { + return op_strict(p, a > b, at, bt); } - -static int -op_ge(int a, int b) -{ - return (a >= b); +static Linetype op_le(long *p, Linetype at, long a, Linetype bt, long b) { + return op_strict(p, a <= b, at, bt); } - -static int -op_eq(int a, int b) -{ - return (a == b); +static Linetype op_ge(long *p, Linetype at, long a, Linetype bt, long b) { + return op_strict(p, a >= b, at, bt); } - -static int -op_ne(int a, int b) -{ - return (a != b); +static Linetype op_eq(long *p, Linetype at, long a, Linetype bt, long b) { + return op_strict(p, a == b, at, bt); } - -static int -op_or(int a, int b) -{ - return (a || b); +static Linetype op_ne(long *p, Linetype at, long a, Linetype bt, long b) { + return op_strict(p, a != b, at, bt); } - -static int -op_and(int a, int b) -{ - return (a && b); +static Linetype op_or(long *p, Linetype at, long a, Linetype bt, long b) { + if (!strictlogic && (at == LT_TRUE || bt == LT_TRUE)) + return (*p = 1, LT_TRUE); + return op_strict(p, a || b, at, bt); +} +static Linetype op_and(long *p, Linetype at, long a, Linetype bt, long b) { + if (!strictlogic && (at == LT_FALSE || bt == LT_FALSE)) + return (*p = 0, LT_FALSE); + return op_strict(p, a && b, at, bt); } /* @@ -823,12 +899,12 @@ op_and(int a, int b) * value of the expression; and (3) a pointer to a char* that points to the * expression to be evaluated and that is updated to the end of the expression * when evaluation is complete. The function returns LT_FALSE if the value of - * the expression is zero, LT_TRUE if it is non-zero, or LT_IF if the - * expression could not be evaluated. + * the expression is zero, LT_TRUE if it is non-zero, LT_IF if the expression + * depends on an unknown symbol, or LT_ERROR if there is a parse failure. */ struct ops; -typedef Linetype eval_fn(const struct ops *, int *, const char **); +typedef Linetype eval_fn(const struct ops *, long *, const char **); static eval_fn eval_table, eval_unary; @@ -839,13 +915,15 @@ static eval_fn eval_table, eval_unary; * element of the table. Innermost expressions have special non-table-driven * handling. */ -static const struct ops { +struct op { + const char *str; + Linetype (*fn)(long *, Linetype, long, Linetype, long); +}; +struct ops { eval_fn *inner; - struct op { - const char *str; - int (*fn)(int, int); - } op[5]; -} eval_ops[] = { + struct op op[5]; +}; +static const struct ops eval_ops[] = { { eval_table, { { "||", op_or } } }, { eval_table, { { "&&", op_and } } }, { eval_table, { { "==", op_eq }, @@ -856,88 +934,119 @@ static const struct ops { { ">", op_gt } } } }; +/* Current operator precedence level */ +static long prec(const struct ops *ops) +{ + return (ops - eval_ops); +} + /* * Function for evaluating the innermost parts of expressions, - * viz. !expr (expr) defined(symbol) symbol number - * We reset the keepthis flag when we find a non-constant subexpression. + * viz. !expr (expr) number defined(symbol) symbol + * We reset the constexpr flag in the last two cases. */ static Linetype -eval_unary(const struct ops *ops, int *valp, const char **cpp) +eval_unary(const struct ops *ops, long *valp, const char **cpp) { const char *cp; char *ep; int sym; + bool defparen; + Linetype lt; cp = skipcomment(*cpp); if (*cp == '!') { - debug("eval%d !", ops - eval_ops); + debug("eval%d !", prec(ops)); cp++; - if (eval_unary(ops, valp, &cp) == LT_IF) - return (LT_IF); - *valp = !*valp; + lt = eval_unary(ops, valp, &cp); + if (lt == LT_ERROR) + return (LT_ERROR); + if (lt != LT_IF) { + *valp = !*valp; + lt = *valp ? LT_TRUE : LT_FALSE; + } } else if (*cp == '(') { cp++; - debug("eval%d (", ops - eval_ops); - if (eval_table(eval_ops, valp, &cp) == LT_IF) - return (LT_IF); + debug("eval%d (", prec(ops)); + lt = eval_table(eval_ops, valp, &cp); + if (lt == LT_ERROR) + return (LT_ERROR); cp = skipcomment(cp); if (*cp++ != ')') - return (LT_IF); + return (LT_ERROR); } else if (isdigit((unsigned char)*cp)) { - debug("eval%d number", ops - eval_ops); + debug("eval%d number", prec(ops)); *valp = strtol(cp, &ep, 0); - cp = skipsym(cp); - } else if (strncmp(cp, "defined", 7) == 0 && endsym(cp[7])) { + if (ep == cp) + return (LT_ERROR); + lt = *valp ? LT_TRUE : LT_FALSE; + cp = ep; + } else if (matchsym("defined", cp) != NULL) { cp = skipcomment(cp+7); - debug("eval%d defined", ops - eval_ops); - if (*cp++ != '(') - return (LT_IF); - cp = skipcomment(cp); - sym = findsym(cp); - if (sym < 0 && !symlist) - return (LT_IF); - *valp = (value[sym] != NULL); - cp = skipsym(cp); + if (*cp == '(') { + cp = skipcomment(cp+1); + defparen = true; + } else { + defparen = false; + } + sym = findsym(&cp); cp = skipcomment(cp); - if (*cp++ != ')') - return (LT_IF); - keepthis = false; + if (defparen && *cp++ != ')') { + debug("eval%d defined missing ')'", prec(ops)); + return (LT_ERROR); + } + if (sym < 0) { + debug("eval%d defined unknown", prec(ops)); + lt = LT_IF; + } else { + debug("eval%d defined %s", prec(ops), symname[sym]); + *valp = (value[sym] != NULL); + lt = *valp ? LT_TRUE : LT_FALSE; + } + constexpr = false; } else if (!endsym(*cp)) { - debug("eval%d symbol", ops - eval_ops); - sym = findsym(cp); - if (sym < 0 && !symlist) - return (LT_IF); - if (value[sym] == NULL) + debug("eval%d symbol", prec(ops)); + sym = findsym(&cp); + if (sym < 0) { + lt = LT_IF; + cp = skipargs(cp); + } else if (value[sym] == NULL) { *valp = 0; - else { + lt = LT_FALSE; + } else { *valp = strtol(value[sym], &ep, 0); if (*ep != '\0' || ep == value[sym]) - return (LT_IF); + return (LT_ERROR); + lt = *valp ? LT_TRUE : LT_FALSE; + cp = skipargs(cp); } - cp = skipsym(cp); - keepthis = false; - } else - return (LT_IF); + constexpr = false; + } else { + debug("eval%d bad expr", prec(ops)); + return (LT_ERROR); + } *cpp = cp; - debug("eval%d = %d", ops - eval_ops, *valp); - return (*valp ? LT_TRUE : LT_FALSE); + debug("eval%d = %d", prec(ops), *valp); + return (lt); } /* * Table-driven evaluation of binary operators. */ static Linetype -eval_table(const struct ops *ops, int *valp, const char **cpp) +eval_table(const struct ops *ops, long *valp, const char **cpp) { const struct op *op; const char *cp; - int val; + long val; + Linetype lt, rt; - debug("eval%d", ops - eval_ops); + debug("eval%d", prec(ops)); cp = *cpp; - if (ops->inner(ops+1, valp, &cp) == LT_IF) - return (LT_IF); + lt = ops->inner(ops+1, valp, &cp); + if (lt == LT_ERROR) + return (LT_ERROR); for (;;) { cp = skipcomment(cp); for (op = ops->op; op->str != NULL; op++) @@ -946,15 +1055,17 @@ eval_table(const struct ops *ops, int *valp, const char **cpp) if (op->str == NULL) break; cp += strlen(op->str); - debug("eval%d %s", ops - eval_ops, op->str); - if (ops->inner(ops+1, &val, &cp) == LT_IF) - return (LT_IF); - *valp = op->fn(*valp, val); + debug("eval%d %s", prec(ops), op->str); + rt = ops->inner(ops+1, &val, &cp); + if (rt == LT_ERROR) + return (LT_ERROR); + lt = op->fn(valp, lt, *valp, rt, val); } *cpp = cp; - debug("eval%d = %d", ops - eval_ops, *valp); - return (*valp ? LT_TRUE : LT_FALSE); + debug("eval%d = %d", prec(ops), *valp); + debug("eval%d lt = %s", prec(ops), linetype_name[lt]); + return (lt); } /* @@ -965,20 +1076,64 @@ eval_table(const struct ops *ops, int *valp, const char **cpp) static Linetype ifeval(const char **cpp) { - int ret; - int val; + Linetype ret; + long val = 0; debug("eval %s", *cpp); - keepthis = killconsts ? false : true; + constexpr = killconsts ? false : true; ret = eval_table(eval_ops, &val, cpp); - return (keepthis ? LT_IF : ret); + debug("eval = %d", val); + return (constexpr ? LT_IF : ret == LT_ERROR ? LT_IF : ret); +} + +/* + * Read a line and examine its initial part to determine if it is a + * preprocessor directive. Returns NULL on EOF, or a pointer to a + * preprocessor directive name, or a pointer to the zero byte at the + * end of the line. + */ +static const char * +skiphash(void) +{ + const char *cp; + + linenum++; + if (fgets(tline, MAXLINE, input) == NULL) { + if (ferror(input)) + err(2, "can't read %s", filename); + else + return (NULL); + } + cp = skipcomment(tline); + if (linestate == LS_START && *cp == '#') { + linestate = LS_HASH; + return (skipcomment(cp + 1)); + } else if (*cp == '\0') { + return (cp); + } else { + return (skipline(cp)); + } +} + +/* + * Mark a line dirty and consume the rest of it, keeping track of the + * lexical state. + */ +static const char * +skipline(const char *cp) +{ + if (*cp != '\0') + linestate = LS_DIRTY; + while (*cp != '\0') + cp = skipcomment(cp + 1); + return (cp); } /* - * Skip over comments and stop at the next character position that is - * not whitespace. Between calls we keep the comment state in the - * global variable incomment, and we also adjust the global variable - * linestate when we see a newline. + * Skip over comments, strings, and character literals and stop at the + * next character position that is not whitespace. Between calls we keep + * the comment state in the global variable incomment, and we also adjust + * the global variable linestate when we see a newline. * XXX: doesn't cope with the buffer splitting inside a state transition. */ static const char * @@ -991,11 +1146,17 @@ skipcomment(const char *cp) return (cp); } while (*cp != '\0') - if (strncmp(cp, "\\\n", 2) == 0) + /* don't reset to LS_START after a line continuation */ + if (strncmp(cp, "\\\r\n", 3) == 0) + cp += 3; + else if (strncmp(cp, "\\\n", 2) == 0) cp += 2; else switch (incomment) { case NO_COMMENT: - if (strncmp(cp, "/\\\n", 3) == 0) { + if (strncmp(cp, "/\\\r\n", 4) == 0) { + incomment = STARTING_COMMENT; + cp += 4; + } else if (strncmp(cp, "/\\\n", 3) == 0) { incomment = STARTING_COMMENT; cp += 3; } else if (strncmp(cp, "/*", 2) == 0) { @@ -1004,10 +1165,18 @@ skipcomment(const char *cp) } else if (strncmp(cp, "//", 2) == 0) { incomment = CXX_COMMENT; cp += 2; + } else if (strncmp(cp, "\'", 1) == 0) { + incomment = CHAR_LITERAL; + linestate = LS_DIRTY; + cp += 1; + } else if (strncmp(cp, "\"", 1) == 0) { + incomment = STRING_LITERAL; + linestate = LS_DIRTY; + cp += 1; } else if (strncmp(cp, "\n", 1) == 0) { linestate = LS_START; cp += 1; - } else if (strchr(" \t", *cp) != NULL) { + } else if (strchr(" \r\t", *cp) != NULL) { cp += 1; } else return (cp); @@ -1019,8 +1188,30 @@ skipcomment(const char *cp) } cp += 1; continue; + case CHAR_LITERAL: + case STRING_LITERAL: + if ((incomment == CHAR_LITERAL && cp[0] == '\'') || + (incomment == STRING_LITERAL && cp[0] == '\"')) { + incomment = NO_COMMENT; + cp += 1; + } else if (cp[0] == '\\') { + if (cp[1] == '\0') + cp += 1; + else + cp += 2; + } else if (strncmp(cp, "\n", 1) == 0) { + if (incomment == CHAR_LITERAL) + error("unterminated char literal"); + else + error("unterminated string literal"); + } else + cp += 1; + continue; case C_COMMENT: - if (strncmp(cp, "*\\\n", 3) == 0) { + if (strncmp(cp, "*\\\r\n", 4) == 0) { + incomment = FINISHING_COMMENT; + cp += 4; + } else if (strncmp(cp, "*\\\n", 3) == 0) { incomment = FINISHING_COMMENT; cp += 3; } else if (strncmp(cp, "*/", 2) == 0) { @@ -1049,12 +1240,36 @@ skipcomment(const char *cp) incomment = C_COMMENT; continue; default: - /* bug */ - abort(); + abort(); /* bug */ } return (cp); } +/* + * Skip macro arguments. + */ +static const char * +skipargs(const char *cp) +{ + const char *ocp = cp; + int level = 0; + cp = skipcomment(cp); + if (*cp != '(') + return (cp); + do { + if (*cp == '(') + level++; + if (*cp == ')') + level--; + cp = skipcomment(cp+1); + } while (level != 0 && *cp != '\0'); + if (level == 0) + return (cp); + else + /* Rewind and re-detect the syntax error later. */ + return (ocp); +} + /* * Skip over an identifier. */ @@ -1066,77 +1281,265 @@ skipsym(const char *cp) return (cp); } +/* + * Skip whitespace and take a copy of any following identifier. + */ +static const char * +getsym(const char **cpp) +{ + const char *cp = *cpp, *sym; + + cp = skipcomment(cp); + cp = skipsym(sym = cp); + if (cp == sym) + return NULL; + *cpp = cp; + return (xstrdup(sym, cp)); +} + +/* + * Check that s (a symbol) matches the start of t, and that the + * following character in t is not a symbol character. Returns a + * pointer to the following character in t if there is a match, + * otherwise NULL. + */ +static const char * +matchsym(const char *s, const char *t) +{ + while (*s != '\0' && *t != '\0') + if (*s != *t) + return (NULL); + else + ++s, ++t; + if (*s == '\0' && endsym(*t)) + return(t); + else + return(NULL); +} + /* * Look for the symbol in the symbol table. If it is found, we return * the symbol table index, else we return -1. */ static int -findsym(const char *str) +findsym(const char **strp) { - const char *cp; + const char *str; int symind; - cp = skipsym(str); - if (cp == str) - return (-1); - if (symlist) - printf("%.*s\n", (int)(cp-str), str); + str = *strp; + *strp = skipsym(str); + if (symlist) { + if (*strp == str) + return (-1); + if (symdepth && firstsym) + printf("%s%3d", zerosyms ? "" : "\n", depth); + firstsym = zerosyms = false; + printf("%s%.*s%s", + symdepth ? " " : "", + (int)(*strp-str), str, + symdepth ? "" : "\n"); + /* we don't care about the value of the symbol */ + return (0); + } for (symind = 0; symind < nsyms; ++symind) { - if (strlcmp(symname[symind], str, cp-str) == 0) { - debug("findsym %s %s", symname[symind], - value[symind] ? value[symind] : ""); + if (matchsym(symname[symind], str) != NULL) { + debugsym("findsym", symind); return (symind); } } return (-1); } +/* + * Resolve indirect symbol values to their final definitions. + */ +static void +indirectsym(void) +{ + const char *cp; + int changed, sym, ind; + + do { + changed = 0; + for (sym = 0; sym < nsyms; ++sym) { + if (value[sym] == NULL) + continue; + cp = value[sym]; + ind = findsym(&cp); + if (ind == -1 || ind == sym || + *cp != '\0' || + value[ind] == NULL || + value[ind] == value[sym]) + continue; + debugsym("indir...", sym); + value[sym] = value[ind]; + debugsym("...ectsym", sym); + changed++; + } + } while (changed); +} + +/* + * Add a symbol to the symbol table, specified with the format sym=val + */ +static void +addsym1(bool ignorethis, bool definethis, char *symval) +{ + const char *sym, *val; + + sym = symval; + val = skipsym(sym); + if (definethis && *val == '=') { + symval[val - sym] = '\0'; + val = val + 1; + } else if (*val == '\0') { + val = definethis ? "1" : NULL; + } else { + usage(); + } + addsym2(ignorethis, sym, val); +} + /* * Add a symbol to the symbol table. */ static void -addsym(bool ignorethis, bool definethis, char *sym) +addsym2(bool ignorethis, const char *sym, const char *val) { + const char *cp = sym; int symind; - char *val; - symind = findsym(sym); + symind = findsym(&cp); if (symind < 0) { if (nsyms >= MAXSYMS) errx(2, "too many symbols"); symind = nsyms++; } - symname[symind] = sym; ignore[symind] = ignorethis; - val = sym + (skipsym(sym) - sym); - if (definethis) { - if (*val == '=') { - value[symind] = val+1; - *val = '\0'; - } else if (*val == '\0') - value[symind] = ""; - else - usage(); + symname[symind] = sym; + value[symind] = val; + debugsym("addsym", symind); +} + +static void +debugsym(const char *why, int symind) +{ + debug("%s %s%c%s", why, symname[symind], + value[symind] ? '=' : ' ', + value[symind] ? value[symind] : "undef"); +} + +/* + * Add symbols to the symbol table from a file containing + * #define and #undef preprocessor directives. + */ +static void +defundefile(const char *fn) +{ + filename = fn; + input = fopen(fn, "rb"); + if (input == NULL) + err(2, "can't open %s", fn); + linenum = 0; + while (defundef()) + ; + if (ferror(input)) + err(2, "can't read %s", filename); + else + fclose(input); + if (incomment) + error("EOF in comment"); +} + +/* + * Read and process one #define or #undef directive + */ +static bool +defundef(void) +{ + const char *cp, *kw, *sym, *val, *end; + + cp = skiphash(); + if (cp == NULL) + return (false); + if (*cp == '\0') + goto done; + /* strip trailing whitespace, and do a fairly rough check to + avoid unsupported multi-line preprocessor directives */ + end = cp + strlen(cp); + while (end > tline && strchr(" \t\n\r", end[-1]) != NULL) + --end; + if (end > tline && end[-1] == '\\') + Eioccc(); + + kw = cp; + if ((cp = matchsym("define", kw)) != NULL) { + sym = getsym(&cp); + if (sym == NULL) + error("missing macro name in #define"); + if (*cp == '(') { + val = "1"; + } else { + cp = skipcomment(cp); + val = (cp < end) ? xstrdup(cp, end) : ""; + } + debug("#define"); + addsym2(false, sym, val); + } else if ((cp = matchsym("undef", kw)) != NULL) { + sym = getsym(&cp); + if (sym == NULL) + error("missing macro name in #undef"); + cp = skipcomment(cp); + debug("#undef"); + addsym2(false, sym, NULL); } else { - if (*val != '\0') - usage(); - value[symind] = NULL; + error("unrecognized preprocessor directive"); } + skipline(cp); +done: + debug("parser line %d state %s comment %s line", linenum, + comment_name[incomment], linestate_name[linestate]); + return (true); } /* - * Compare s with n characters of t. - * The same as strncmp() except that it checks that s[n] == '\0'. + * Concatenate two strings into new memory, checking for failure. */ -static int -strlcmp(const char *s, const char *t, size_t n) +static char * +astrcat(const char *s1, const char *s2) { - while (n-- && *t != '\0') - if (*s != *t) - return ((unsigned char)*s - (unsigned char)*t); - else - ++s, ++t; - return ((unsigned char)*s); + char *s; + int len; + size_t size; + + len = snprintf(NULL, 0, "%s%s", s1, s2); + if (len < 0) + err(2, "snprintf"); + size = (size_t)len + 1; + s = (char *)malloc(size); + if (s == NULL) + err(2, "malloc"); + snprintf(s, size, "%s%s", s1, s2); + return (s); +} + +/* + * Duplicate a segment of a string, checking for failure. + */ +static const char * +xstrdup(const char *start, const char *end) +{ + size_t n; + char *s; + + if (end < start) abort(); /* bug */ + n = (size_t)(end - start) + 1; + s = malloc(n); + if (s == NULL) + err(2, "malloc"); + snprintf(s, n, "%s", start); + return (s); } /* @@ -1162,5 +1565,6 @@ error(const char *msg) else warnx("%s: %d: %s (#if line %d depth %d)", filename, linenum, msg, stifline[depth], depth); + closeio(); errx(2, "output may be truncated"); } diff --git a/usr.bin/unifdef/unifdef.h b/usr.bin/unifdef/unifdef.h new file mode 100644 index 00000000000..0cd5c8c5b05 --- /dev/null +++ b/usr.bin/unifdef/unifdef.h @@ -0,0 +1,50 @@ +/* + * Copyright (c) 2012 - 2013 Tony Finch + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + */ + +#include + +#include +#include +#include +#include +#include +#include +#include +#include + +/* portabiity stubs */ + +#define fbinmode(fp) (fp) + +#define replace(old,new) rename(old,new) + +static FILE * +mktempmode(char *tmp, int mode) +{ + int fd = mkstemp(tmp); + if (fd < 0) return (NULL); + fchmod(fd, mode & (S_IRWXU|S_IRWXG|S_IRWXO)); + return (fdopen(fd, "wb")); +} diff --git a/usr.bin/unifdef/version.h b/usr.bin/unifdef/version.h new file mode 100644 index 00000000000..f34ccb81b21 --- /dev/null +++ b/usr.bin/unifdef/version.h @@ -0,0 +1,2 @@ +"@(#) $Version: unifdef-2.10 $\n" +"@(#) $Date: 2014/04/18 21:11:34 $\n"